WEBVTT

1
00:00:00.192 --> 00:00:01.638
I just heard gunshots outside.

2
00:00:01.920 --> 00:00:02.281
Is this PHP?

3
00:00:05.081 --> 00:00:06.162
Open code got hacked.

4
00:00:06.222 --> 00:00:08.203
Is that why these sirens are outside?

5
00:00:08.243 --> 00:00:11.405
Cybersecurity experts are calling in Black October.

6
00:00:12.085 --> 00:00:14.926
I think the branding of Black October is a little dramatic.

7
00:00:14.966 --> 00:00:19.849
All of these things ended up being not as big a deal as they kind of were advertised to be.

8
00:00:19.869 --> 00:00:29.394
E. Mudra, Certificate of Authority, during their key generation ceremony, I wish I had some candles and crystals and then we could meditate a little bit.

9
00:00:29.534 --> 00:00:32.415
Generate the key, make it bigger than the number four.

10
00:00:32.616 --> 00:00:33.176
Now...

11
00:00:33.476 --> 00:00:42.466
We have a completely unrelated story, not even a week later, where another private key having to do with PLS is the number four again.

12
00:00:42.666 --> 00:00:46.050
Four has become the six seven of cybersecurity, which is so funny.

13
00:00:46.170 --> 00:00:47.191
What the hell happened?

14
00:00:53.244 --> 00:00:54.846
Good morning, hackers.

15
00:00:54.946 --> 00:00:59.350
Welcome back to the best podcast on the internet for hackers, The Lowdown, presented by Mays.

16
00:00:59.730 --> 00:01:02.233
We are unfortunately no longer in Las Vegas.

17
00:01:02.253 --> 00:01:05.837
You're back to my basement and Matt's office, but we had a great time in Vegas.

18
00:01:05.877 --> 00:01:08.779
Matt, how was your OneCon experience, man?

19
00:01:08.799 --> 00:01:09.060
How'd it go?

20
00:01:09.444 --> 00:01:15.110
The only unfortunate part about not being in Las Vegas anymore for me personally is those sick cameras.

21
00:01:15.130 --> 00:01:17.673
We got footage of the episode that came out.

22
00:01:17.894 --> 00:01:21.958
Yes, Sentinel-1 had like a $10,000 podcast set up.

23
00:01:22.318 --> 00:01:23.260
That's not what I have.

24
00:01:24.752 --> 00:01:25.913
I was so mad at this guy.

25
00:01:25.953 --> 00:01:27.453
I was like, this guy's going to cost me $20,000.

26
00:01:27.734 --> 00:01:34.237
And I was like, no, Matt, the video footage is not why your YouTube channel isn't popping off.

27
00:01:34.277 --> 00:01:36.438
You know, like it's you, it's a skill issue.

28
00:01:36.478 --> 00:01:38.439
It's not that it's not an 8k anyway.

29
00:01:38.499 --> 00:01:39.920
Yeah, no, we had a good time, man.

30
00:01:40.460 --> 00:01:41.221
Really impressed.

31
00:01:41.461 --> 00:01:43.502
Never been to a conference at the ARIA conference center.

32
00:01:43.522 --> 00:01:44.362
Yep.

33
00:01:44.602 --> 00:01:45.463
Nice conference center.

34
00:01:46.103 --> 00:01:48.444
Sentinel one put on a tight show, man.

35
00:01:48.564 --> 00:01:53.507
Like they had really, we popped down to their like university breakout sessions.

36
00:01:54.247 --> 00:02:02.634
On the lower level that first day, I looked at all of the titles of all the different talks and not even like computer low level, like basement.

37
00:02:03.895 --> 00:02:13.883
So, but pop down all of the titles of all the tracks that were going on were like exactly what should have been happening this week, which is usually hard to do when you pull off a conference to like have timely...

38
00:02:14.523 --> 00:02:17.585
like sessions, you know, because you're planning it so far in advance.

39
00:02:17.605 --> 00:02:17.885
Right.

40
00:02:17.925 --> 00:02:28.750
Like no, no, they did really good to like break it out into like AI offense, AI defense, like securing AI, like, you know, all that kind of stuff was really, really good.

41
00:02:29.251 --> 00:02:34.533
And it seemed like we popped our head into a bunch of them and it seemed like they were really doing hands on lab type stuff.

42
00:02:34.553 --> 00:02:35.374
So that was awesome.

43
00:02:35.594 --> 00:02:38.655
We wanted to go check it out, but like every room we went into was very quiet.

44
00:02:38.695 --> 00:02:39.856
Everyone was like deep in the weeds.

45
00:02:39.896 --> 00:02:42.877
It's like, all right, I don't want to interrupt people likely in their flow state.

46
00:02:42.917 --> 00:02:43.157
Right.

47
00:02:43.177 --> 00:02:44.418
So no, no, thanks.

48
00:02:44.458 --> 00:02:45.799
But it looked like they were having a good time.

49
00:02:45.819 --> 00:02:46.399
So it was cool.

50
00:02:46.419 --> 00:02:51.021
And then the the eSport thing was just I blew my expectations out of the water.

51
00:02:51.161 --> 00:02:56.063
You know, I'm thinking like live hacking competition, generally not very fun to watch.

52
00:02:56.143 --> 00:02:57.064
They nailed it.

53
00:02:57.144 --> 00:02:57.704
They nailed it.

54
00:02:58.784 --> 00:03:01.346
Go watch last week's episode if you have not.

55
00:03:01.406 --> 00:03:06.248
We interviewed the winner of that competition and you can kind of see some footage from the con in there.

56
00:03:06.908 --> 00:03:09.349
And we should probably put out some, like we took some B-roll and stuff.

57
00:03:09.369 --> 00:03:15.112
Like we should probably just put out some more onto social, like onto our social channels, just so people get the vibe of the conference overall.

58
00:03:15.858 --> 00:03:16.418
Good job.

59
00:03:16.738 --> 00:03:17.439
Yeah, it was a great time.

60
00:03:17.619 --> 00:03:21.200
Yeah, it takes a big personality to make competitive hacking actually interesting.

61
00:03:21.220 --> 00:03:24.201
I've been in like those competitions and they are not fun to watch.

62
00:03:24.662 --> 00:03:29.324
Yeah, but you know what happened like while we were in Vegas?

63
00:03:29.624 --> 00:03:31.484
Cybersecurity Awareness Month.

64
00:03:32.405 --> 00:03:33.625
Listen up, everybody.

65
00:03:33.885 --> 00:03:38.487
We heard that you were trying to not be cybersecurity aware and this is what happens.

66
00:03:38.527 --> 00:03:39.988
This is what happens when you don't listen.

67
00:03:40.448 --> 00:03:44.211
When you try to not be cybersecurity aware, this is what we have to do.

68
00:03:44.291 --> 00:03:47.153
We have to rebrand it to Black October.

69
00:03:47.753 --> 00:03:48.073
Yeah.

70
00:03:48.214 --> 00:04:06.747
So, guys, if you weren't aware, right, like, basically the day we landed in Las Vegas, there were, like, three or four, like, major industry-disrupting disclosures regarding, like, PHP, Node.js, OpenCode got hacked, a certificate authority had the private key four on some other intermediates, like...

71
00:04:07.347 --> 00:04:07.747
Not great.

72
00:04:08.187 --> 00:04:18.333
That being said, I think the branding of Black October is a little dramatic, and we'll go into why all of these things ended up being not as big a deal as they kind of were advertised to be.

73
00:04:18.793 --> 00:04:23.596
Still not good, obviously, but this isn't like, you know, the Black Tuesday for cyber.

74
00:04:23.616 --> 00:04:24.857
Like, we're kind of pinning it up to be.

75
00:04:24.977 --> 00:04:25.897
You know what it's giving?

76
00:04:25.917 --> 00:04:26.698
You know what it's giving?

77
00:04:26.838 --> 00:04:27.178
What's that?

78
00:04:28.018 --> 00:04:45.945
Mythos It's giving mythos It's giving mythos It does play a role We will go into that But like It ain't that big kid It ain't that big chief Dude it has a Wikipedia page It must be big It must be real It must be really bad The vulnerability has a website Therefore it must be legit guys That's how this works

79
00:04:47.185 --> 00:04:57.269
To date myself really quick, I've been writing blogs and recording content about breaches weekly going back to 2009.

80
00:04:57.789 --> 00:04:58.149
Weekly.

81
00:04:58.229 --> 00:04:58.489
Wow.

82
00:04:58.509 --> 00:04:58.989
2009.

83
00:04:59.149 --> 00:05:01.470
I don't want to know where you are again.

84
00:05:01.490 --> 00:05:03.811
I don't want to know.

85
00:05:04.091 --> 00:05:04.891
Freshman in high school.

86
00:05:05.091 --> 00:05:05.351
Okay.

87
00:05:05.411 --> 00:05:05.672
Okay.

88
00:05:05.832 --> 00:05:06.512
Okay.

89
00:05:06.792 --> 00:05:10.153
But yeah, I like wrote for this blog, LiquidMationSecurityDigest.

90
00:05:10.233 --> 00:05:17.536
I actually, I did a daily news roundup on that blog of like links because we didn't have, you know, Daily is bonkers in 09.

91
00:05:17.576 --> 00:05:18.756
That's fucking crazy.

92
00:05:18.776 --> 00:05:19.156
Yeah, dude.

93
00:05:19.176 --> 00:05:23.238
I did like on a blog because that's how we all got our news was RSS feeders and blogs.

94
00:05:23.318 --> 00:05:26.879
And I would like do like a roundup of all the different security news of the day.

95
00:05:27.459 --> 00:05:28.300
breaches were part of it.

96
00:05:28.400 --> 00:05:31.141
Then we started a podcast for Liquid Matrix.

97
00:05:31.701 --> 00:05:32.842
I couldn't even tell you what year.

98
00:05:32.922 --> 00:05:34.382
It was before podcasts were cool.

99
00:05:35.303 --> 00:05:41.886
It was when the name podcast still made sense because you were listening to it on an iPad, iPod.

100
00:05:42.146 --> 00:05:43.867
And we did over a hundred episodes of that.

101
00:05:43.947 --> 00:05:46.968
And we had a section in our show notes called breaches.

102
00:05:47.208 --> 00:05:50.470
And so every week when we did record that podcast, we talked about breaches.

103
00:05:51.030 --> 00:05:52.191
This is not new.

104
00:05:52.491 --> 00:05:53.252
This is not new.

105
00:05:53.832 --> 00:05:55.454
A bunch of vulnerabilities came out.

106
00:05:55.474 --> 00:05:58.136
A bunch of really crazy vulnerabilities came out this week.

107
00:05:58.336 --> 00:05:59.216
They got patches.

108
00:05:59.717 --> 00:05:59.917
Yeah.

109
00:06:00.598 --> 00:06:01.758
End of news cycle.

110
00:06:01.818 --> 00:06:04.040
Like, I mean, there's stuff to talk about.

111
00:06:04.080 --> 00:06:05.201
Like, the industry is changing.

112
00:06:05.301 --> 00:06:07.102
AI is changing the speed of things.

113
00:06:07.143 --> 00:06:07.903
Like, got it.

114
00:06:08.183 --> 00:06:08.524
But...

115
00:06:09.252 --> 00:06:13.454
Like, to say that, you know, the floor is falling out from this stuff.

116
00:06:13.834 --> 00:06:21.258
The Japanese prime minister declared a state of national emergency and told people to stockpile cans of food and water because of PHP vulnerabilities.

117
00:06:21.858 --> 00:06:30.923
I got DMs of people screenshotting their emergency alerts going off on their phone because, like, police activity down the block or whatever, like, stay inside.

118
00:06:30.943 --> 00:06:34.425
I got DMs saying, is this why I'm getting this alert?

119
00:06:34.545 --> 00:06:34.805
Yeah.

120
00:06:35.025 --> 00:06:37.248
Is there police activity around the corner because of PHP?

121
00:06:37.268 --> 00:06:38.350
Because of PHP.

122
00:06:38.430 --> 00:06:40.032
That's crazy, dude.

123
00:06:40.072 --> 00:06:41.634
We have missed it.

124
00:06:41.694 --> 00:06:42.916
We have jumped the shark.

125
00:06:42.976 --> 00:06:44.898
We have missed the plot.

126
00:06:45.159 --> 00:06:46.621
I just heard gunshots outside.

127
00:06:46.901 --> 00:06:47.282
Is this PHP?

128
00:06:47.302 --> 00:06:48.043
Like, what the fuck?

129
00:06:50.863 --> 00:06:51.964
Open code got hacked.

130
00:06:52.004 --> 00:06:53.985
Is that why these sirens are outside?

131
00:06:54.005 --> 00:06:57.066
Dude, it's absolutely bananas.

132
00:06:57.126 --> 00:06:57.987
We have lost it.

133
00:06:58.027 --> 00:06:59.668
We've completely lost the plot.

134
00:07:00.128 --> 00:07:03.329
So I do want to talk about some PHP stuff.

135
00:07:03.349 --> 00:07:03.749
Bring it back.

136
00:07:03.809 --> 00:07:08.172
You did your PHP video live from your hotel room.

137
00:07:08.632 --> 00:07:09.032
Baller.

138
00:07:10.052 --> 00:07:10.573
Respect.

139
00:07:10.873 --> 00:07:13.474
I didn't bring my YouTube set up to Vegas.

140
00:07:14.575 --> 00:07:19.517
I did mine yesterday to kind of do the zoom out like aftermath type video.

141
00:07:20.377 --> 00:07:26.639
We talked about PHP stuff as we knew it live in Vegas on the podcast that's already out from last week.

142
00:07:28.120 --> 00:07:39.284
I think it's really valuable to look at this as a case study in what the hell is going on in this industry and the reaction and all of this kind of stuff and exploitation.

143
00:07:39.704 --> 00:07:42.945
Did you see this live tracker that Census put out?

144
00:07:43.726 --> 00:07:44.826
No, I don't think so.

145
00:07:45.346 --> 00:07:45.666
All right.

146
00:07:46.486 --> 00:07:47.667
Live tracker of what?

147
00:07:48.347 --> 00:07:53.789
Okay, so this is the PHP 2-Shell live global dashboard.

148
00:07:53.849 --> 00:07:55.949
Basically their version of a PewPew map.

149
00:07:56.230 --> 00:08:01.311
If you guys don't know Sensus, they're like a Shodan competitor, I guess.

150
00:08:02.411 --> 00:08:10.554
There's a few of these companies out there that scan the whole internet and kind of like make everything that they find available via their API and dashboard and whatever they find.

151
00:08:10.594 --> 00:08:14.155
And it's almost like a search engine for just devices on the internet, right?

152
00:08:14.656 --> 00:08:21.376
And so, of course, census can then fingerprint vulnerable versions of PHP that are exposed on the internet.

153
00:08:22.223 --> 00:08:25.705
So they've got this great dashboard, php2shell.census.com.

154
00:08:25.885 --> 00:08:27.806
So super easy dashboard you can pull up.

155
00:08:28.526 --> 00:08:29.967
Fairly live tracking.

156
00:08:30.227 --> 00:08:32.568
I think they said they could be like a few hours behind.

157
00:08:33.749 --> 00:08:39.072
Fairly live tracking of vulnerable versus patched PHP servers live on the internet.

158
00:08:39.812 --> 00:08:48.957
At its peak, like when the bug came out October 21st, or I guess when the patch came out, because I think the bug leaked the 20th, right?

159
00:08:49.437 --> 00:08:49.878
Mm-hmm.

160
00:08:50.778 --> 00:08:59.944
when the patch came out the 21st, there was 43 million, 200 something thousand vulnerable PHP servers on the thing.

161
00:09:00.024 --> 00:09:02.406
It's a big effing number when we're talking about these things.

162
00:09:03.426 --> 00:09:13.593
Usually headline making vulns, if you guys don't track vulns for a living like we do, headline making ones could be in the tens of thousands of like things and people would still be freaking out.

163
00:09:14.013 --> 00:09:15.914
So 43 million is a very, very big number.

164
00:09:15.934 --> 00:09:16.574
That's a couple.

165
00:09:16.594 --> 00:09:17.134
That's a few.

166
00:09:17.474 --> 00:09:17.674
Yeah.

167
00:09:18.514 --> 00:09:27.576
The first two days after I saw this tracker go live, the internet at large was patching 4.3 million servers a day.

168
00:09:27.736 --> 00:09:30.117
So the first two days, 4.3 million, 4.3 million.

169
00:09:30.137 --> 00:09:35.618
I was like, okay, if this is telling us anything about the pace of patching, we should be pretty good.

170
00:09:35.798 --> 00:09:37.959
Obviously there'll be like a long tail or whatever.

171
00:09:38.479 --> 00:09:41.362
That third day, boom, 300,000 or something like that.

172
00:09:41.402 --> 00:09:43.103
What was it?

173
00:09:43.123 --> 00:09:46.566
600,000, then 300,000, then 200,000 and 300,000.

174
00:09:46.606 --> 00:09:55.453
So we've like gone from 4.3 million to a few hundred thousand patches per day, which is still, like I said, 10,000 vulnerable servers would have been a headline making news story.

175
00:09:55.473 --> 00:09:56.975
We're patching hundreds of thousands a day.

176
00:09:57.255 --> 00:09:57.555
Cool.

177
00:09:57.835 --> 00:10:05.342
We still have 32 million servers to get through as a collective internet to like worry about with this volume out there.

178
00:10:05.362 --> 00:10:06.322
Yeah, that's still a lot.

179
00:10:06.343 --> 00:10:06.663
Yeah.

180
00:10:07.003 --> 00:10:07.263
Yeah.

181
00:10:07.343 --> 00:10:10.004
So I thought that that 4 million pace was going to be pretty cool.

182
00:10:10.564 --> 00:10:12.444
It's not, right?

183
00:10:13.004 --> 00:10:20.746
The saving grace here is, and I kind of want to pick your brain about this too, because you know a little bit about the exploit and you've kind of looked into it.

184
00:10:21.546 --> 00:10:24.627
Why aren't we seeing as much active exploitation about this?

185
00:10:24.747 --> 00:10:28.608
I kind of want to talk a little bit before I pass it over to you.

186
00:10:28.628 --> 00:10:29.308
Yeah.

187
00:10:29.468 --> 00:10:29.808
We...

188
00:10:30.932 --> 00:10:37.315
we saw a lot of like POC type stuff come out with the vuln announcement, right?

189
00:10:37.696 --> 00:10:41.278
So I was like, this is going to be bad because we're seeing POCs, the patch isn't out yet.

190
00:10:41.678 --> 00:10:45.940
It turns out a lot of those POCs were AI generated slop, hallucinations.

191
00:10:46.040 --> 00:10:47.621
The models thought they had an exploit.

192
00:10:47.721 --> 00:10:58.307
So the researcher thought they had an exploit and the researcher or the model weren't good enough or smart enough to realize that they did not have like a working POC is really what a lot of those videos were.

193
00:10:59.547 --> 00:11:00.328
That being said,

194
00:11:01.352 --> 00:11:02.513
the original researcher.

195
00:11:02.793 --> 00:11:05.935
And then now I'm seeing a number of other researchers.

196
00:11:05.975 --> 00:11:15.740
I think some of the first two I saw was like Volncheck and Trail of Bits came out and said that they have production grade, like working exploits on a bunch of different versions of stuff.

197
00:11:16.000 --> 00:11:17.621
So it is possible, right?

198
00:11:17.641 --> 00:11:18.261
It is possible.

199
00:11:18.341 --> 00:11:28.487
Even with ASLR, even with all sorts of protections, the exploitation is possible even at scale, it seems to like write, if they're calling it a production grade POC.

200
00:11:29.367 --> 00:11:34.789
Um, we haven't, we were not seeing any confirmed active, active exploitation for a few days.

201
00:11:35.029 --> 00:11:43.931
Then the first word I saw of it, and I need to apologize, uh, you know, say sorry, basically to everyone on this.

202
00:11:44.391 --> 00:11:45.152
I amplified it.

203
00:11:45.352 --> 00:11:55.615
Unit 42 from Palo Alto networks, who I think is one of the better threat intelligence teams on the planet, uh, put out that they saw active exploitation of this, right?

204
00:11:56.255 --> 00:11:56.975
Uh, a few days ago.

205
00:11:57.375 --> 00:11:58.756
on the 24th.

206
00:11:59.197 --> 00:11:59.477
Okay.

207
00:11:59.677 --> 00:12:02.159
So this is still just like three days after patch comes out.

208
00:12:02.840 --> 00:12:04.582
Unit 42 says active exploitation.

209
00:12:04.942 --> 00:12:09.966
The article was worded a little bit weird and you can even see my tweets go like some people were responding to me.

210
00:12:09.986 --> 00:12:14.090
So like I reposted unit 42 researchers saying active exploitation.

211
00:12:14.110 --> 00:12:16.072
I was like, there we go guys, active exploitation.

212
00:12:16.132 --> 00:12:16.893
I trust these people.

213
00:12:17.693 --> 00:12:21.836
And people were responding like, that article seems a little weird.

214
00:12:21.856 --> 00:12:22.957
Seems like it doesn't say that.

215
00:12:23.537 --> 00:12:24.758
And I responded, I was like, you're right.

216
00:12:24.858 --> 00:12:29.822
It is worded weird, but I'm just kind of trusting that unit 42 is unit 42.

217
00:12:30.342 --> 00:12:31.663
They have since walked this back.

218
00:12:32.664 --> 00:12:33.284
They were wrong.

219
00:12:33.665 --> 00:12:37.047
So I got to talk to some people, try to figure it out.

220
00:12:37.207 --> 00:12:37.608
Seems like,

221
00:12:38.428 --> 00:12:43.913
like a team there saw a PHP server be exploited.

222
00:12:44.133 --> 00:12:47.816
And then they were going through a bunch of post exploitation stuff in that blog.

223
00:12:48.456 --> 00:12:52.540
And they just kind of jumped to say, oh, well this must've been PHP to shell.

224
00:12:53.192 --> 00:12:53.492
Right.

225
00:12:53.753 --> 00:12:54.133
And whatever.

226
00:12:54.173 --> 00:12:55.614
And they, and they've walked it back.

227
00:12:55.954 --> 00:12:59.417
So we've had like a lot of misinformation about active exploitation.

228
00:12:59.777 --> 00:13:02.760
Now we, we, we have some that seems pretty legit, right?

229
00:13:02.840 --> 00:13:05.942
That like exploit is happening now we're recording this on the 28th.

230
00:13:06.183 --> 00:13:10.706
So like, seems like in the last few days it's starting to happen, but I I'm surprised we're not seeing more.

231
00:13:10.746 --> 00:13:12.147
What's, what's your thoughts on this whole thing?

232
00:13:12.708 --> 00:13:13.889
Yeah, I guess.

233
00:13:13.909 --> 00:13:21.155
So to set the stage, the trend we're going to kind of highlight in today's podcast is like over since like over, um,

234
00:13:22.476 --> 00:13:26.101
central centralization over over sensationalism, I guess is the word.

235
00:13:26.582 --> 00:13:30.587
It's kind of a trend that's happening right now because security is like so in the zeitgeist.

236
00:13:30.747 --> 00:13:32.509
AI is making security the hot topic.

237
00:13:32.890 --> 00:13:40.379
So when a thing comes out like an exploit in PHP that affects 70 percent of the globe, it's going to catch a lot of wind.

238
00:13:41.100 --> 00:13:47.728
The problem with this vulnerability is that it is an extremely complex memory exploit, right?

239
00:13:48.229 --> 00:13:55.617
So I'm not going to cover it too deep in detail because we talked about it in the previous episode, and I want to repeat myself, but the PHP to shell exploit,

240
00:13:56.338 --> 00:14:13.105
is a use after free vulnerability because of PHP's use of the long jump function, which basically is a way in C of if you encounter an error state, you can use long jump to return to a saved context, a saved state that you were previously in.

241
00:14:13.125 --> 00:14:15.566
So it's like, okay, we set the state here.

242
00:14:15.926 --> 00:14:17.967
We go into a function that is fallible, right?

243
00:14:18.067 --> 00:14:20.388
If it fails, we call long jump return over here.

244
00:14:20.968 --> 00:14:27.131
That returns the program's instruction pointer, stack pointer, and registers all back to where it was previously.

245
00:14:27.871 --> 00:14:36.635
The problem is the RAM and the other global variables of the function or of the running program are the way that they were after the failure, right?

246
00:14:36.655 --> 00:14:42.117
So if you don't properly account for all of the ways a program can fail, you can have some logical issues that happen here.

247
00:14:42.497 --> 00:14:56.663
The PHP to shell bug was basically when you are initializing a session during a file upload, if that file upload fails and causes a PHP allocator to go out of memory, the session is still partially initialized, but it's freed.

248
00:14:56.923 --> 00:15:00.024
But then a further on request uses that freed session.

249
00:15:00.044 --> 00:15:06.527
So you can use that vulnerability to do an arbitrary read, arbitrary write, bypass ASLR, and do your code execution.

250
00:15:07.267 --> 00:15:25.742
The problem with that, though, is that a use after free, most of the time, depends on a very precise grooming of the heap so that when your next allocation comes in to fill that freed slot, you know what structure it is and you know that you control that structure and at what offsets you control that structure.

251
00:15:26.382 --> 00:15:33.347
The issue is PHP is such a low-level, name of the channel, by the way, thing in the stack of the web frameworks, right?

252
00:15:33.428 --> 00:15:35.889
WordPress, Drupal, Joomla, whatever, um,

253
00:15:36.610 --> 00:15:53.327
And every unique implementation of PHP, or not implementation, rather, every unique usage of PHP, and all of them can be behind an Apache 2 reverse proxy, an NGINX reverse proxy that are all going to mangle the requests in different ways.

254
00:15:54.028 --> 00:15:55.630
Those change automatically.

255
00:15:55.975 --> 00:15:59.156
the memory shape, the heap shape of the exploit.

256
00:15:59.756 --> 00:16:02.897
On top of that, you have different versions of PHP.

257
00:16:02.917 --> 00:16:07.538
You have different versions of the frameworks that all can depend on different versions of PHP.

258
00:16:07.838 --> 00:16:13.919
You have this graph explosion of basically the different kinds of memory states that your program can be in.

259
00:16:14.479 --> 00:16:16.920
And it's important to remember too that hackers aren't clairvoyant.

260
00:16:16.960 --> 00:16:18.940
Like, you know, sometimes we like to think we are, right?

261
00:16:19.240 --> 00:16:20.321
But we don't know.

262
00:16:20.481 --> 00:16:23.361
Like maybe we can fingerprint your WordPress and your WordPress is WordPress.

263
00:16:23.421 --> 00:16:24.542
I'm making up a number, 7-Eleven.

264
00:16:24.562 --> 00:16:25.562
I have no idea if that's real or not.

265
00:16:26.222 --> 00:16:35.269
But that WordPress 7.11 can be riding on PHP 8.2 and making it more numbers, which also was reverse proxied by Nginx 4.3.

266
00:16:35.309 --> 00:16:36.070
Again, more fake numbers.

267
00:16:36.970 --> 00:16:45.097
So that combination of version numbers causes a unique heap shape that is not just universally fingerprintable.

268
00:16:46.071 --> 00:16:46.911
I'm sorry, Matt, you were talking?

269
00:16:46.931 --> 00:16:47.051
What?

270
00:16:47.191 --> 00:16:47.452
No.

271
00:16:47.512 --> 00:16:48.992
Hey, man, those are all real numbers.

272
00:16:49.072 --> 00:16:49.272
Okay.

273
00:16:49.352 --> 00:16:51.033
Three and four and three and four.

274
00:16:51.073 --> 00:16:51.593
Okay.

275
00:16:51.613 --> 00:16:53.574
To make sure I'm not saying numbers that aren't real.

276
00:16:53.674 --> 00:16:54.934
Okay.

277
00:16:55.295 --> 00:16:57.395
So ultimately, the bug is real.

278
00:16:57.435 --> 00:16:58.256
The bug is bad.

279
00:16:58.456 --> 00:17:04.278
But A, it takes a lot of information that may not be publicly disclosable to reliably exploit it.

280
00:17:04.838 --> 00:17:07.099
And two, the bug is hard.

281
00:17:07.119 --> 00:17:11.661
So even if you have all of those pieces of information, it's not necessarily just a free...

282
00:17:12.281 --> 00:17:13.262
Command injection, right?

283
00:17:13.342 --> 00:17:23.613
Even in the Kalief leak that we saw the video of him, the person that found the bug running the exploit, it said attempt number one of eight, because it's not just like a get in, get out.

284
00:17:23.673 --> 00:17:29.380
You have to sometimes crash the daemon and hope the heap is in the right shape before you can get in.

285
00:17:29.820 --> 00:17:31.061
And that's assuming there's no other traffic, right?

286
00:17:31.081 --> 00:17:34.183
Because, again, I'm not sure about the worker to heap layout.

287
00:17:34.523 --> 00:17:40.867
But it may be the case that you are sharing a worker with another person on the Internet talking to that server, and they're also manipulating the heap.

288
00:17:41.267 --> 00:17:45.269
So it's like if you don't lock that worker to just you, that also influences things.

289
00:17:45.709 --> 00:17:49.772
So real bug, real scary, but not real easy to exploit.

290
00:17:50.573 --> 00:17:59.539
Yeah, and there's a lot of debates going on right now of if, since it's so hard to exploit, is this as serious as we were all making it seem?

291
00:17:59.559 --> 00:18:02.801
And I think the consensus is like, well, yes, right?

292
00:18:02.821 --> 00:18:06.003
Like, yeah, because 32 million vulnerable things, right?

293
00:18:06.423 --> 00:18:16.450
And because now we're starting to see some of this actual weaponization of this bug to be used as the initial access vector for complex intrusions, right?

294
00:18:16.590 --> 00:18:17.370
We have an example.

295
00:18:17.390 --> 00:18:21.952
I didn't want to share this because I got a little gun shy on the unit 42 thing.

296
00:18:21.992 --> 00:18:22.272
Right.

297
00:18:22.352 --> 00:18:25.733
I was like, oh, I shared something from someone I trusted and I got burnt.

298
00:18:26.333 --> 00:18:28.474
And this is from a group that I do not know.

299
00:18:28.714 --> 00:18:28.954
Right.

300
00:18:29.054 --> 00:18:32.935
It seems like this is for some research team over at dream dream group.

301
00:18:33.255 --> 00:18:36.017
Looks like they do like sovereign AI for governments.

302
00:18:36.297 --> 00:18:36.657
Got it.

303
00:18:36.697 --> 00:18:43.781
So didn't know, not familiar with the team, but I do have some, again, some other people I trust that said this is legit.

304
00:18:43.901 --> 00:18:58.650
So like willing to talk about it now, but it seems like we do have some evidence of chained vulnerabilities in the first in the chain being PHP to shell exploit, rooting 25 servers, including US government servers is the title of the article.

305
00:18:58.730 --> 00:19:00.671
We read through this research.

306
00:19:01.011 --> 00:19:01.512
Wait, wait, wait.

307
00:19:02.588 --> 00:19:20.874
Ed, you said that you're not clairvoyant, but just last week, in the example when we were talking about how pointing the electron super vuln-finding microscope at PHP found more vulns, the example you used was something called dirty frag, where that happened.

308
00:19:21.534 --> 00:19:26.696
You were like, yeah, it was kind of like dirty frag, where it came out and we found a whole bunch of other things like it.

309
00:19:27.405 --> 00:19:28.526
I thought you weren't clairvoyant.

310
00:19:28.887 --> 00:19:32.270
That is the volume that was used in one of those variances.

311
00:19:32.290 --> 00:19:37.375
It was used to chain in this exploit that we're seeing against an unnamed firewall vendor.

312
00:19:37.756 --> 00:19:44.142
We have seen some headlines of some firewall vendors that are running PHP, maybe with or without ASLR sometimes, too.

313
00:19:44.162 --> 00:19:45.223
A lot of stuff.

314
00:19:45.303 --> 00:19:48.627
Don't know if that has anything to do with what we're talking about here.

315
00:19:48.647 --> 00:19:49.247
Citrix?

316
00:19:50.568 --> 00:19:53.470
Citrix or Palo, unfortunately, right?

317
00:19:53.850 --> 00:19:56.772
So in different versions of some stuff.

318
00:19:57.172 --> 00:20:02.134
Obviously, as we keep talking about, we might as well have a section on the podcast called Edge Device Breach of the Week, right?

319
00:20:02.154 --> 00:20:04.035
These are highly targeted things.

320
00:20:04.055 --> 00:20:04.816
They're deployed.

321
00:20:04.876 --> 00:20:05.456
The code is...

322
00:20:07.277 --> 00:20:10.519
you know, written in C. I mean, there's just tons of stuff, right?

323
00:20:11.219 --> 00:20:13.380
Reasons why they're highly targeted.

324
00:20:13.660 --> 00:20:26.387
So this one, because of all that memory stuff that you just outlined, it seems like this exploit has some techniques here to do some memory exhaustion to force a predictable memory layout.

325
00:20:26.407 --> 00:20:29.188
But you said that that may or may not be accurate.

326
00:20:29.988 --> 00:20:41.978
on this research, but at least they are acknowledging that that is part of the trickiness of exploiting, but it seems like this weaponization had an answer to that and this chain.

327
00:20:42.058 --> 00:20:43.759
Can you go into this research a little bit?

328
00:20:44.219 --> 00:20:48.903
Yes, I was reading this, and so they said predictable memory layout, right?

329
00:20:50.418 --> 00:20:51.899
That could mean one of two things.

330
00:20:51.919 --> 00:21:01.662
When you say predictable memory layout, you're trying to effectively bypass two issues in the way that allocators work that tend to get in the way of exploitation, right?

331
00:21:02.163 --> 00:21:05.264
The first one is, you know, address space layout randomization.

332
00:21:05.344 --> 00:21:10.246
So basically, if an attacker gets control of a pointer or the return address...

333
00:21:11.214 --> 00:21:13.074
The question is, where do they point it to, right?

334
00:21:13.094 --> 00:21:19.336
If you want to do an arbitrary read or arbitrary write, you have to put a known good value in that location to do the hacking.

335
00:21:19.916 --> 00:21:28.878
What ASLR does is it puts the base address of a page, a new allocation by the kernel, at a random location so that you don't know what memory address to fill in that location.

336
00:21:30.018 --> 00:21:37.259
What they could be asserting here is that they're allocating a ton of stuff to reduce the overall entropy of ASLR, right?

337
00:21:37.339 --> 00:21:39.140
ASLR only has an entropy of...

338
00:21:40.009 --> 00:21:40.889
Oh, God, what is it?

339
00:21:40.969 --> 00:21:46.932
It's like you're missing the bottom 16 bits, and you're missing the top 16 bits.

340
00:21:46.952 --> 00:21:49.453
You get total addressable entropy of, like, I think it's 32 bits.

341
00:21:49.493 --> 00:21:51.214
So in 64-bit land, 32 bits.

342
00:21:51.834 --> 00:21:55.075
In 32-bit land, it's, like, much less.

343
00:21:55.095 --> 00:21:56.496
You get, like, maybe four or five bits of entropy.

344
00:21:56.516 --> 00:22:01.018
It could be them trying to bypass ASLR by reducing the available page address range.

345
00:22:01.718 --> 00:22:05.800
I think what's actually going on here is this is them doing the heap grooming I was talking about, right?

346
00:22:05.840 --> 00:22:07.541
So when you're doing a heap grooming,

347
00:22:08.301 --> 00:22:18.784
You are trying to throw enough data at the program where you are hitting allocations that you know are going to occur to cause the heap shape to be a certain way.

348
00:22:18.804 --> 00:22:25.806
What you're doing is creating what is called a hole in the heap so that you know when your next allocation occurs, it's going to land in this exact location.

349
00:22:26.626 --> 00:22:33.835
Um, this is done so that you know what structure you will control in the use after free case where you have the pointer that points to the freed thing.

350
00:22:34.156 --> 00:22:36.438
You need to put data there that you control in a certain shape.

351
00:22:36.478 --> 00:22:42.286
And so this is them, in my opinion, um, using large amounts of data to heap the groom to maintain that control.

352
00:22:42.886 --> 00:22:49.815
I don't see a world where they're attacking a device where these pieces of data would enable them to bypass ASLR.

353
00:22:49.855 --> 00:22:51.457
It just doesn't make sense to me.

354
00:22:51.537 --> 00:22:54.220
Either that, or like we talked about before, they're landing on a firewall.

355
00:22:54.641 --> 00:22:58.225
This could be a firewall that just doesn't have ASLR enabled, right?

356
00:22:58.245 --> 00:23:00.348
A lot of embedded devices, even in the year of our Lord, 2026, um...

357
00:23:02.170 --> 00:23:02.770
just don't have it on.

358
00:23:03.010 --> 00:23:05.031
I think we've seen Citrix doesn't have it on.

359
00:23:05.431 --> 00:23:08.992
The Palo Alto VMs don't have it on, but the hardware devices do.

360
00:23:09.372 --> 00:23:13.034
A lot of devices that run MIPS just like, oops, don't enable ASLR.

361
00:23:13.054 --> 00:23:13.514
I don't know why.

362
00:23:14.514 --> 00:23:19.796
So in my opinion, the chances of this being an ASLR bypass are low, but very likely heat grooming.

363
00:23:20.016 --> 00:23:22.717
Or it could just be them hiding their exploit in chaff, right?

364
00:23:22.737 --> 00:23:25.858
They could be trying to obfuscate where their shit lies in the actual exploit itself.

365
00:23:25.878 --> 00:23:26.178
So there's a

366
00:23:27.122 --> 00:23:30.123
You know, a couple different things that could be going on here.

367
00:23:30.143 --> 00:23:32.543
There's different TTPs, but who's to know?

368
00:23:32.943 --> 00:23:33.983
So super interesting.

369
00:23:34.103 --> 00:23:41.165
It does seem like this research may or may not have gotten on the attackers kind of like operational boxes, right?

370
00:23:41.205 --> 00:23:46.546
Because they're seeming to have some information about the actual campaign that you wouldn't get from just the compromised boxes.

371
00:23:47.126 --> 00:23:47.946
So super interesting.

372
00:23:48.006 --> 00:23:54.908
The other thing that I thought was kind of crazy, because you don't see this a whole lot, and especially crazy because it's PHP to shell, is...

373
00:23:55.608 --> 00:24:02.093
is the initial access vector to the firewall is p2shell plus the Fragnesia thing, that dirty frag variant, right?

374
00:24:03.114 --> 00:24:10.079
Then from the firewall to continue the exploit into the network, they use p2shell again.

375
00:24:10.853 --> 00:24:11.954
onto the next box.

376
00:24:12.294 --> 00:24:13.636
So they literally use it twice.

377
00:24:14.236 --> 00:24:16.378
They use it once with the frag thing.

378
00:24:16.638 --> 00:24:20.221
Then they use it again to get onto a management server and root that.

379
00:24:20.301 --> 00:24:26.807
So even not your internet exposed PHP things, we now have proof.

380
00:24:27.207 --> 00:24:34.213
Like your internal PHP type stuff is now that's being used to get onto that box.

381
00:24:34.733 --> 00:24:37.776
And then they pivoted all the way to domain controllers via some LDAP bug.

382
00:24:38.036 --> 00:24:44.419
I want to highlight, though, I tend to dunk on embedded companies because their software tends to be not very good.

383
00:24:44.759 --> 00:24:52.583
But what I will notice that I didn't think about until just now, they had to use Fragnesia, which means that the web server is not running as Root.

384
00:24:53.183 --> 00:24:53.723
We did it.

385
00:24:53.903 --> 00:24:57.104
We did proper operating system enforced user control.

386
00:24:57.144 --> 00:24:57.904
Holy shit.

387
00:24:58.004 --> 00:24:59.204
It only took 26 years.

388
00:24:59.805 --> 00:25:00.985
So that's kind of cool, right?

389
00:25:01.045 --> 00:25:08.927
It's ironically that the firewall got rooted by an actual LPE, but the server landed as root.

390
00:25:09.147 --> 00:25:10.647
Completely backwards.

391
00:25:10.927 --> 00:25:12.207
Yeah, yeah.

392
00:25:12.227 --> 00:25:22.830
So they got on the server and apparently they had a stored LDAP credential and they decrypted the LDAP credential, which gave them the ability to land as the system on the DC, which is not great.

393
00:25:23.330 --> 00:25:23.550
Yeah.

394
00:25:23.870 --> 00:25:25.991
So I think that it's time.

395
00:25:26.011 --> 00:25:36.634
I think it's time to fly in an expert into this episode of the lowdown and talk about Black October and what we're seeing and how this is all changing in AI exploitation land.

396
00:25:36.654 --> 00:25:37.134
What do you think?

397
00:25:37.154 --> 00:25:37.735
You think it's time?

398
00:25:37.755 --> 00:25:38.255
Let's do it, man.

399
00:25:38.275 --> 00:25:38.635
Bring it in.

400
00:25:45.379 --> 00:25:46.640
Harry, welcome to the podcast.

401
00:25:46.660 --> 00:25:47.721
Appreciate you coming on, man.

402
00:25:48.102 --> 00:25:51.865
You know, pretty boring news week, as I'm sure you've heard.

403
00:25:51.925 --> 00:25:54.328
Not like really a lot going on in the cybersecurity space.

404
00:25:54.768 --> 00:26:00.173
Before we get into all these bugs and what AI is doing to the industry, can you kind of kick us off with an introduction?

405
00:26:00.213 --> 00:26:00.674
What do you do?

406
00:26:00.734 --> 00:26:01.194
Who are you?

407
00:26:01.214 --> 00:26:02.375
And what's your shtick?

408
00:26:03.655 --> 00:26:04.176
Yeah, of course.

409
00:26:04.296 --> 00:26:05.297
Yeah, good to be here, guys.

410
00:26:05.617 --> 00:26:06.277
Yeah, I'm Harry.

411
00:26:06.358 --> 00:26:08.019
I'm co-founder and CEO of Maze.

412
00:26:08.580 --> 00:26:11.722
For those who don't know, we're a new application security company, pretty much.

413
00:26:11.783 --> 00:26:23.113
So, as you can imagine, using AI to figure out in code, where are the vulnerabilities, either understanding a CVE, right, like of the thousands that you might have, which of them actually matter to you and how do you fix them?

414
00:26:23.353 --> 00:26:24.795
And then we're also using the same...

415
00:26:25.315 --> 00:26:27.457
technology to go and find vulnerabilities where we didn't know about.

416
00:26:27.577 --> 00:26:32.300
So like what kind of people are trying to do with Mythos, but a lot cheaper and a lot more efficient effectively.

417
00:26:32.481 --> 00:26:39.546
So we do a wide range of things, a bit like Sneak or that kind of space in the past, but a modern take on the same idea.

418
00:26:40.066 --> 00:26:40.287
Nice.

419
00:26:40.547 --> 00:26:40.747
Awesome.

420
00:26:40.767 --> 00:26:47.492
That's a great space to be in right now because as I'm sure you're aware, there are a couple bugs, one or two, not a ton.

421
00:26:47.532 --> 00:26:50.635
I spent the last two weeks on Twitter

422
00:26:51.315 --> 00:26:57.847
Seeing people go, like people outside of security usually, but they're like, man, someone should really build an AI product to solve this problem.

423
00:26:57.967 --> 00:27:00.652
And I'm just there scrolling like, yeah, I'm here.

424
00:27:01.513 --> 00:27:02.815
Hey, don't mind me.

425
00:27:03.432 --> 00:27:03.792
Yeah.

426
00:27:03.832 --> 00:27:10.176
So one of my favorite questions to kind of kick off these kinds of conversations, I've talked to some other founders in your space, right?

427
00:27:10.276 --> 00:27:11.096
There's a few.

428
00:27:11.576 --> 00:27:14.858
If you go to the black hat floor, the booths are getting crowded.

429
00:27:14.958 --> 00:27:21.682
And a few years ago, there was like some mumblings of AI, then everyone kind of had to have a story and now everyone's an AI company, right?

430
00:27:22.202 --> 00:27:23.904
You guys were an early mover in the space.

431
00:27:24.424 --> 00:27:25.725
What are you kind of seeing?

432
00:27:25.745 --> 00:27:28.448
Like, why is AI getting so good at finding bugs?

433
00:27:28.468 --> 00:27:36.815
And how are you differentiating from like security company that is doing this for a living versus like what the frontier models can do out of the box?

434
00:27:37.375 --> 00:27:38.295
Yeah, yeah, yeah.

435
00:27:38.495 --> 00:27:38.775
It's true.

436
00:27:38.795 --> 00:27:42.776
Like, if you go around Black Hat or anything similar now, all the big companies have chosen one of two lanes.

437
00:27:43.376 --> 00:27:46.957
Either they're pretending to be an AI company since they were founded, you know, 20 years ago.

438
00:27:47.757 --> 00:27:52.518
Or if they can't figure out a way to sell that story, they just say, we were built to secure AI.

439
00:27:52.838 --> 00:27:54.919
Like, that's, everyone is one of those two lanes.

440
00:27:55.339 --> 00:27:56.319
I like that.

441
00:27:56.359 --> 00:27:59.380
So it's, like, tricky for us, I guess, sometimes to stand out.

442
00:27:59.420 --> 00:27:59.760
But, yeah.

443
00:28:00.840 --> 00:28:05.901
Yeah, I guess, so I was working on machine learning products before LLMs came out.

444
00:28:05.921 --> 00:28:08.042
So kind of come from that old, older world.

445
00:28:08.522 --> 00:28:13.243
And I guess as a company, what we've tried to do is say AI agents are ridiculously powerful.

446
00:28:13.263 --> 00:28:18.885
Even very early versions of them back in like 23, 24 were clearly going to completely change our security work.

447
00:28:18.905 --> 00:28:23.046
Because you could finally have this thing that could go and crawl through all the contexts that you might care about.

448
00:28:23.606 --> 00:28:28.409
Every security tool until then basically was limited by not understanding the full context around a situation, right?

449
00:28:28.429 --> 00:28:39.296
It was screaming at you, let's say, about a PHP vulnerability that you had in your environment and tell you that it's the most urgent thing in the world, not knowing that there was a bunch of context around it that meant it wasn't exploitable or similar.

450
00:28:39.977 --> 00:28:44.900
But the, I guess, premise of what we're trying to do is to say those agents are also really hard to get right at scale.

451
00:28:45.400 --> 00:28:48.963
So if I want to run one once on my own machine, I might get an interesting result once.

452
00:28:49.223 --> 00:28:49.563
Yeah.

453
00:28:49.683 --> 00:28:55.986
But if I want to scan every line of code, every piece of cloud infrastructure, I want to do it every single day for the rest of time.

454
00:28:56.406 --> 00:28:58.987
And I want to make sure that my results are almost 100% accurate.

455
00:28:59.507 --> 00:29:04.109
And I want them to be affordable in some moderate sense for an enterprise.

456
00:29:04.149 --> 00:29:05.630
That is a really hard challenge.

457
00:29:05.670 --> 00:29:12.152
And so we just lean in headfirst into that challenge and say, there's actually a ton of software you need to build around the models to make the most of that challenge.

458
00:29:12.172 --> 00:29:13.333
And so that's what we specialize in.

459
00:29:13.933 --> 00:29:14.133
Nice.

460
00:29:14.514 --> 00:29:18.337
So it's not, hey, Mythos, find me a zero day, make no mistakes.

461
00:29:18.438 --> 00:29:20.440
It's actually a lot of tooling that goes into this.

462
00:29:20.740 --> 00:29:21.240
Yeah, exactly.

463
00:29:21.260 --> 00:29:23.743
Well, loads, yeah, and they can find impressive things doing that.

464
00:29:24.123 --> 00:29:28.388
But when you compare, like, basically what we find is like every month that you put into building...

465
00:29:29.008 --> 00:29:53.163
software and ways of handling the data crucially around the models, you basically get like step changes in both accuracy and cost efficiency, the more the more kind of built around them, which is why always makes me laugh when people just like hook up a latest front end model and say, great, we do security now because yeah, you do and you'll find interesting stuff, but it'll be like riddled with false positives, it'll be different results day after day, and it'll be astronomically expensive most of the time.

466
00:29:54.436 --> 00:30:00.041
Was it Tom or Steve, Matt, that was telling us that their board directed them to use a frontier model?

467
00:30:00.101 --> 00:30:04.205
They heard investors basically telling companies, make sure we're using frontier models.

468
00:30:04.225 --> 00:30:05.747
And it's like, okay, and?

469
00:30:05.907 --> 00:30:06.688
Like, what else?

470
00:30:06.768 --> 00:30:08.650
What would you like me to do with this?

471
00:30:08.750 --> 00:30:09.871
It's such a buzzword right now.

472
00:30:11.073 --> 00:30:11.934
It wasn't their board.

473
00:30:11.974 --> 00:30:12.754
It was their customers.

474
00:30:12.774 --> 00:30:18.677
So they were running into it, but their customers' boards were mandating frontier model usage.

475
00:30:18.837 --> 00:30:33.364
And the quote that he said that's just been rattling around in my head is that they've never seen a board mandate MFA, but now they're seeing the board's mandate model usage to stay on top of cybersecurity type things.

476
00:30:33.404 --> 00:30:34.825
And I was like, yeah, that's a really good point.

477
00:30:34.945 --> 00:30:36.886
And the headlines are really driving a lot of the narrative here.

478
00:30:37.788 --> 00:30:43.654
There's a lot of conversation right now about where this goes as the limit approaches infinity, right?

479
00:30:43.875 --> 00:30:46.878
As time goes on, where does software go?

480
00:30:46.918 --> 00:30:52.744
There's kind of the one attacker mindset, which is like we just find all the bugs and it's the Wild West for the rest of eternity.

481
00:30:53.064 --> 00:30:59.071
But then the logical counter to that is like, well, we also find all the bugs and the bugs go away and software is secure and we live in utopia.

482
00:30:59.091 --> 00:30:59.191
Yeah.

483
00:31:00.914 --> 00:31:02.056
Where do you think this runs, right?

484
00:31:02.116 --> 00:31:09.911
Do we live in a more secure software environment, or do we just go back to the 90s for the rest of eternity, right?

485
00:31:09.931 --> 00:31:10.532
How does this land?

486
00:31:12.018 --> 00:31:12.979
Yeah, it's interesting.

487
00:31:13.259 --> 00:31:20.966
I'm sure you guys saw, there was a bit of chatter online recently about this, this idea of like volume zero, which is this idea that we're just going to eventually eradicate all the bugs.

488
00:31:21.166 --> 00:31:25.671
You know, we'll just use LLMs to go back and improve all of, all of like history software.

489
00:31:25.971 --> 00:31:27.652
And then going forward, we're just not going to write any new ones.

490
00:31:28.393 --> 00:31:31.135
And there's, there's like a, there's a little bit of truth in that, right?

491
00:31:31.156 --> 00:31:33.077
Because coding agents are clearly getting so much better.

492
00:31:34.295 --> 00:31:37.417
Certain classes of vulnerabilities, I think we probably can almost eradicate.

493
00:31:37.817 --> 00:31:43.320
But at the same time, like what problem in security have we ever solved like fully?

494
00:31:44.100 --> 00:31:51.264
Usually zero of them because there's always like you're always fighting this kind of like actor on the other side that basically is just using the same technology you are.

495
00:31:51.885 --> 00:31:54.686
And so it's a bit like, I guess, like turning up to.

496
00:31:54.706 --> 00:31:55.507
Yeah.

497
00:31:56.487 --> 00:32:03.276
you know, the first person to invent the gun and they turn up to like a sword fight and they're like, right, well, I've won all fights forever, ever more now.

498
00:32:03.536 --> 00:32:03.697
Sure.

499
00:32:03.837 --> 00:32:10.506
And not realizing that like the next wave of people are going to come back at them with guns and the next way we're going to come back with FFV, F15s and whatever it might be.

500
00:32:10.586 --> 00:32:13.430
So I think we're going to find ourselves in a similar spot where, uh,

501
00:32:14.615 --> 00:32:20.616
basically the types of vulnerabilities that we all have languishing our backlogs today, some of those might go away.

502
00:32:22.036 --> 00:32:27.618
But logically then what you would see happening is the models are just going to be just as good on the attacking side.

503
00:32:27.638 --> 00:32:31.278
The models are good enough to write perfectly secure code and secure all our code perfectly.

504
00:32:31.778 --> 00:32:35.739
Then they need to be incredibly good at distributing and generating attacks.

505
00:32:36.099 --> 00:32:39.260
Those attacks are then orders of magnitude more complex than the ones we understand today.

506
00:32:40.000 --> 00:32:44.264
And the thing that my co-founder Santiago always says is like, there is no such thing as perfectly secure code.

507
00:32:44.705 --> 00:32:47.387
There's like perfectly secure, like snippets of code, right?

508
00:32:47.407 --> 00:32:57.497
Like an individual piece of code can be in theory perfectly support secure, but put it in a sufficiently complex system and you can never perfectly secure it because you don't know what all the interactions in that system are going to look like.

509
00:32:57.897 --> 00:33:00.980
And so the idea that we're just going to spit out loads of perfectly secure code

510
00:33:01.621 --> 00:33:08.285
kind of fails to acknowledge the fact that actually the really hard to find vulnerabilities are these really subtle things that pop up in big complex systems.

511
00:33:08.746 --> 00:33:12.588
And there's just no way that we're gonna be able to, you know, completely eradicate them.

512
00:33:12.628 --> 00:33:15.430
So sadly, probably we're gonna solve some of the problem.

513
00:33:15.830 --> 00:33:17.431
We're gonna get a new problem pop up somewhere else.

514
00:33:17.651 --> 00:33:20.393
And you know, everything's gonna probably feel the same as depressing as that is.

515
00:33:20.723 --> 00:33:31.314
I think this transitions really nicely into what's turning into one of our favorite questions to ask people around like the AI attacker versus defender asymmetry, right?

516
00:33:31.375 --> 00:33:36.740
And I know like I've always really hated the mantra of like defenders have to be right all the time.

517
00:33:36.780 --> 00:33:40.384
Attackers only have to be right once because like it's not really true when you pull the thread.

518
00:33:41.285 --> 00:33:44.687
Yeah, they only need to be right once to get in, but then they can get caught immediately.

519
00:33:44.987 --> 00:33:49.230
Like, you know, there's zero days are not invisibility cloaks, all that kind of stuff, right?

520
00:33:49.290 --> 00:34:02.878
But when it comes to this conversation that we're having, right, it seems like right now we are in an AI attacker is beating AI defender, power defender, finger on the scale of the asymmetry, right?

521
00:34:04.259 --> 00:34:07.861
there's definitely first mover advantage to the AI attacker, right?

522
00:34:07.961 --> 00:34:16.805
And less bureaucracy in their organizations to be able to use these new things to get better at finding whatever it is that they're finding.

523
00:34:17.586 --> 00:34:25.970
You, as someone who's on the, like trying to help the AI-powered defender side, how do you see kind of this asymmetry playing out like long-term?

524
00:34:26.191 --> 00:34:30.333
Is this like forever in the attacker's advantage or how does this go?

525
00:34:31.514 --> 00:34:31.914
Good question.

526
00:34:31.934 --> 00:34:35.878
Yeah, I agree with you on the fact that the whole asymmetry thing doesn't quite stack up.

527
00:34:35.998 --> 00:34:38.521
Because the other thing is the defenders have all the data as well.

528
00:34:39.001 --> 00:34:40.703
Attackers are trying to get in blind.

529
00:34:41.544 --> 00:34:42.244
Right, right, right.

530
00:34:42.304 --> 00:34:43.906
All the information on their own environment.

531
00:34:43.946 --> 00:34:45.527
So there's a massive asymmetry there.

532
00:34:46.368 --> 00:34:50.472
So last week, on Tuesday, Wednesday, I was writing a blog post.

533
00:34:50.592 --> 00:34:53.154
And the blog post was all about where are all the AI attacks?

534
00:34:53.415 --> 00:34:53.615
Yeah.

535
00:34:53.806 --> 00:34:56.007
And it was all about how there's not actually been that many of them.

536
00:34:56.107 --> 00:34:56.807
Where are they?

537
00:34:56.827 --> 00:34:57.447
What's going on?

538
00:34:58.228 --> 00:34:59.768
And then obviously all the news here.

539
00:34:59.788 --> 00:35:03.089
And I was like, I think this is maybe not the time to post this.

540
00:35:04.290 --> 00:35:11.012
But in that research that I did, there's been like very few cases of like documented AI-based attacks.

541
00:35:11.032 --> 00:35:13.733
So someone clearly using an LM or a series of agents.

542
00:35:13.753 --> 00:35:16.855
There's been a few more in the last few weeks, as you guys have probably seen.

543
00:35:17.415 --> 00:35:19.936
But I compared it to Felony Bench.

544
00:35:20.616 --> 00:35:25.941
The thing that shows how many felonies OpenAI and Anthropic and everyone has committed.

545
00:35:26.261 --> 00:35:28.202
Yeah, so good.

546
00:35:28.222 --> 00:35:39.932
I did my best job of trying to find all the cases of real AI-based attacks in the world, and I counted them, and then I counted all the items on FelonyBench across OpenAI, Anthropic, Google, etc.

547
00:35:39.992 --> 00:35:46.498
There's 23 felonies by the actual AI labs, and there's 23 confirmed cases of actual AI-based attacks

548
00:35:46.878 --> 00:35:50.380
So like the accidents and the real stuff is head to head, apparently.

549
00:35:50.420 --> 00:35:51.660
Obviously, that always hides stuff.

550
00:35:51.680 --> 00:35:54.381
There's always way more, you know, looking at the surface that we don't know about.

551
00:35:54.461 --> 00:35:55.982
But it is kind of weird.

552
00:35:56.002 --> 00:35:57.663
We actually haven't seen that many yet.

553
00:35:58.243 --> 00:36:06.987
The only thing I can rely on here is that maybe attackers actually, in most cases, are not as numerous or as sophisticated as some of us like to think.

554
00:36:07.427 --> 00:36:13.009
You know, some of us like to think that maybe, like, you give them enough power and we're just going to have, like, endless attacks around the world at all times.

555
00:36:13.510 --> 00:36:15.090
But there's still a lot of work required.

556
00:36:16.851 --> 00:36:24.293
Like initial access, you can automate exploiting a phone, let's say you can automate maybe even doing a bit of a bit of moving around a certain environment.

557
00:36:24.574 --> 00:36:26.794
There's still a load more work left to do on top of that.

558
00:36:27.034 --> 00:36:32.036
Like, let's say you want to extort someone, you still need to contact that entity.

559
00:36:32.056 --> 00:36:34.037
You need to figure out how to extort them out of some money.

560
00:36:34.357 --> 00:36:39.879
You need to take the criminal risk of going through that process and not getting found, you need to communicate within your own network.

561
00:36:39.899 --> 00:36:40.459
Like there's still...

562
00:36:40.959 --> 00:36:47.962
Just because AI makes the attack a bit easier doesn't make the whole situation just effortless is maybe one argument for why we haven't seen a bit more of it.

563
00:36:48.442 --> 00:36:53.985
But yeah, I think that asymmetry, the more negative case on the asymmetry is it actually just takes them a while to build the tools.

564
00:36:54.760 --> 00:36:59.004
Like in my past life before Mays, I used to work on phishing.

565
00:36:59.065 --> 00:37:02.068
So trying to stop, you know, like really advanced phishing attacks, basically.

566
00:37:02.248 --> 00:37:08.334
And what you saw there was that the tech to do pretty advanced phishing at scale was available way earlier than you saw it at scale.

567
00:37:08.835 --> 00:37:14.641
And there was this kind of like wave that happened where effectively people went and they realized that you could spear phish through phishing.

568
00:37:14.861 --> 00:37:16.041
like email gateways pretty easily.

569
00:37:16.261 --> 00:37:18.182
And they'd realized a way of doing that pretty easily.

570
00:37:18.662 --> 00:37:20.442
And they realized a way you could get money from that pretty easily.

571
00:37:20.582 --> 00:37:22.643
And then you saw like a very, very low level of it.

572
00:37:23.123 --> 00:37:30.925
And then you saw phishing as a service kits kind of start to come out and you saw like, you know, $2, you know, uh, just put it in a domain and we'll just cover it for you.

573
00:37:31.045 --> 00:37:33.286
And then you saw this massive spike of that kind of activity.

574
00:37:33.946 --> 00:37:42.268
And I kind of feel we might be in that phase at the moment where basically it takes a while for you to kind of package the tech up and easy to use things that like anyone on.

575
00:37:42.628 --> 00:37:43.990
the dark web can kind of go and use.

576
00:37:44.591 --> 00:37:48.756
And we might be about to see that asymmetry start to kind of actually play through a bit more.

577
00:37:48.836 --> 00:37:50.157
But it's really hard to know.

578
00:37:50.177 --> 00:37:52.140
And it's obviously hard to see because we don't see a lot of the data.

579
00:37:52.537 --> 00:37:53.337
It's a really good point.

580
00:37:53.417 --> 00:37:56.759
We just saw this week a hacker get caught.

581
00:37:57.799 --> 00:38:02.922
And I guess, yeah, they were only active for a few weeks and only made $210.

582
00:38:03.102 --> 00:38:10.125
But they had stolen like a bunch of records and then just like no OPSEC, no actual follow through on the payment or anything like that.

583
00:38:10.405 --> 00:38:13.026
They're like VC-backed companies with no revenue.

584
00:38:13.226 --> 00:38:14.687
They're like, yeah, we're growing really fast.

585
00:38:14.767 --> 00:38:16.167
You still got monetized what you're doing.

586
00:38:17.748 --> 00:38:23.298
So the AI-powered hacker got in, stole some records, but then was pretty bad at being a criminal.

587
00:38:23.399 --> 00:38:23.779
You're right.

588
00:38:24.340 --> 00:38:25.903
We're starting to see evidence of that for sure.

589
00:38:25.923 --> 00:38:26.785
It's funny.

590
00:38:28.013 --> 00:38:28.874
Yeah, that's really cool.

591
00:38:28.894 --> 00:38:32.456
You know, the more people that we ask that, the more different answers we get.

592
00:38:32.516 --> 00:38:34.617
It's like why it's turning into one of our favorite questions.

593
00:38:34.637 --> 00:38:45.803
Because I think a lot of people just at large, if you were to just poll the audience, right, in the industry, I think a lot of people would lean the AI attacker advantage, just long term, that would be their answer, right?

594
00:38:46.663 --> 00:38:51.726
But then, yeah, we've talked to a few people who are like in the weeds with customers actually like going up against this.

595
00:38:51.766 --> 00:38:55.668
And they're like, no, I think the defender long term is going to have the advantage here.

596
00:38:55.968 --> 00:38:57.410
Yeah, the defender has a massive advantage.

597
00:38:57.430 --> 00:39:03.397
The asymmetry used to come from the fact that defenders found it very hard to understand their own environment, right?

598
00:39:03.417 --> 00:39:05.879
It was like we had very rudimentary ways of understanding our own environments.

599
00:39:06.420 --> 00:39:10.445
Therefore, the attacker from the outside had the asymmetry of you just need to be right once, as you say.

600
00:39:11.065 --> 00:39:14.810
And the defender had some access to their internal data that gave them a bit of advantage in response.

601
00:39:14.850 --> 00:39:16.351
And then you kind of, you know, you have to find it out.

602
00:39:16.972 --> 00:39:21.033
We now have increasingly almost perfect understanding of our internal environments.

603
00:39:21.073 --> 00:39:21.253
Right.

604
00:39:21.633 --> 00:39:29.056
Uh, we have things that can read every line of code in our company for us and understand everything about it, how it all fits together, where the risks might be.

605
00:39:29.456 --> 00:39:31.576
We can run like red teaming exercises.

606
00:39:32.677 --> 00:39:33.737
Like, uh.

607
00:39:34.407 --> 00:39:37.510
essentially ongoing all the time trying to find out what could go wrong.

608
00:39:38.111 --> 00:39:43.295
We can inventory what we've got in our environments where all the problems that used to hold us back internally have got a lot easier.

609
00:39:43.836 --> 00:39:45.317
Attackers still don't have all that data.

610
00:39:45.337 --> 00:39:53.065
They have to find ways to hack us from the outside using essentially the same techniques, but sped up and made a little bit more sophisticated.

611
00:39:53.085 --> 00:39:53.985
So I think it's still going to be

612
00:39:54.951 --> 00:39:59.815
Difficult for defenders, but theoretically, AI gives defenders more of an advantage than attackers.

613
00:40:00.055 --> 00:40:05.919
We were talking earlier, kind of dunking on this whole Black Friday label, right?

614
00:40:06.500 --> 00:40:08.281
Or Black October, whatever it is, right?

615
00:40:08.361 --> 00:40:12.104
Everyone's kind of looking at the last couple of weeks in cybersecurity and being like, we've never seen anything like this.

616
00:40:12.124 --> 00:40:13.104
And it's like, we kind of have.

617
00:40:13.365 --> 00:40:15.026
And we've kind of seen a whole lot worse, right?

618
00:40:15.066 --> 00:40:18.688
Like, you know, WannaCry, Log4J, people got hacked, right?

619
00:40:18.729 --> 00:40:20.410
Like, nothing has really kind of come of this.

620
00:40:21.110 --> 00:40:28.451
Do you have any just thoughts looking at the last week about the industry, PHP, the reaction to everything that we're seeing?

621
00:40:28.611 --> 00:40:30.812
What does it look like from your point of view?

622
00:40:30.832 --> 00:40:34.492
We're kind of starting to look a bit ridiculous, aren't we?

623
00:40:34.512 --> 00:40:43.714
It looks like the fire alarm went off and we all ran out of the building screaming and now we're all stood outside looking at the building still not on fire and like, ah, this is kind of embarrassing us.

624
00:40:45.654 --> 00:40:46.294
But yeah, you're right.

625
00:40:46.334 --> 00:40:49.135
I think the interesting thing about last week was...

626
00:40:49.895 --> 00:40:52.538
It's coming at a time of a lot of just AI doomerism in general.

627
00:40:52.718 --> 00:40:54.760
So it's just like plugging straight into that.

628
00:40:55.521 --> 00:41:06.112
And also the nature of the vulnerabilities, they were almost, if you were writing a script and you were trying to write a series of vulnerabilities that would scare the hell out of everyone relative to their actual impact,

629
00:41:06.733 --> 00:41:20.164
You would probably go after PHP, which is one of the most widely used things that could have a vulnerability, but its ratio of how widely used it is relative to how at risk it might be for key things around the world is pretty low.

630
00:41:20.665 --> 00:41:24.488
You would maybe go after TLS, just because it scares everyone like the internet is vulnerable.

631
00:41:25.789 --> 00:41:29.212
And you would probably set up the vulnerabilities in such a way that a bunch of

632
00:41:29.872 --> 00:41:34.344
very visible online websites took their websites down for a while to scare, you know, to, to scare everyone.

633
00:41:34.384 --> 00:41:34.565
Right.

634
00:41:34.585 --> 00:41:36.751
So it was kind of like, I think,

635
00:41:37.746 --> 00:41:43.971
it was almost the perfect storm of what would be the scariest potential set of things happening relative to their actual risk.

636
00:41:44.451 --> 00:41:53.578
And then now still, like I think I was looking up yesterday, I think the two known breaches of PHP, like the major ones that came out from, I think Palo Alto or someone else, they've both been rescinded.

637
00:41:53.598 --> 00:41:55.520
And it was like, yeah, it actually wasn't anything to do with it.

638
00:41:56.881 --> 00:42:01.344
And I saw something, I think it was Grey Noise today had some honey pots up for the last like week.

639
00:42:01.765 --> 00:42:04.487
And they've just about managed to find one of them get exploited.

640
00:42:04.507 --> 00:42:04.707
Yeah.

641
00:42:05.187 --> 00:42:07.068
Yeah, by the same threat actor, right?

642
00:42:07.088 --> 00:42:10.650
So the number of threat actors actually being skilled enough to wield it is pretty low.

643
00:42:10.810 --> 00:42:11.550
Yeah, exactly.

644
00:42:11.751 --> 00:42:12.131
Exactly.

645
00:42:12.191 --> 00:42:19.835
So, yeah, I don't know what that means for us, because the other side of the equation I fear a little bit is when, like, Mythos happened, and Anthropic obviously just...

646
00:42:20.565 --> 00:42:23.027
like ramped up the scare factor to a thousand.

647
00:42:23.308 --> 00:42:34.498
And then when this happens, and again, like people involved, just ramp up the scare factor ridiculously and everyone just, it becomes this big echo chamber and everyone outside of security is, I don't know about you guys, but I just got friends, you know, messaging me be like, Hey, are we all screwed?

648
00:42:34.778 --> 00:42:34.898
Um,

649
00:42:35.819 --> 00:42:47.008
And so now that that's happened a few times, the next time, let's say something that is a bit more like Log4j or maybe worse than that happens, and everyone starts going crazy, a lot of people are going to be like, oh, it's just the next scare.

650
00:42:47.588 --> 00:42:48.889
And they're not going to actually respond to it.

651
00:42:48.909 --> 00:42:52.032
They're maybe not going to put the actions in place so they can respond quickly the next time.

652
00:42:52.472 --> 00:43:05.422
So we're in this quite dangerous loop where actually the overreaction is creating a lot of, I think, for naturally skeptical security people, as many are, it's kind of creating this loop where everyone's kind of turning off to the actual problem.

653
00:43:05.942 --> 00:43:07.803
And that might come back to buy us, unfortunately.

654
00:43:08.644 --> 00:43:09.464
That's a really good point.

655
00:43:09.704 --> 00:43:11.105
The boy who cried four.

656
00:43:13.166 --> 00:43:15.787
Harry, what's your go-to, what's your stack right now, man?

657
00:43:15.887 --> 00:43:21.611
What's your, you got a local model running at home, or you went in an anthropic homie, what's your stack right now?

658
00:43:21.671 --> 00:43:27.734
Me, personally, I mean, I'm the founder of the CEO, so I have bored, my like day-to-day work is ridiculously boring, right?

659
00:43:27.794 --> 00:43:33.557
It's like Slack and Claude and Granola and a few other things.

660
00:43:34.507 --> 00:43:37.190
Internally though at Maze, we use a little bit of everything.

661
00:43:37.230 --> 00:43:47.881
We are like the model company's worst nightmares, I guess, because we have endless amounts of infrastructure to monitor very quickly how good the models are, both accuracy and cost efficiency.

662
00:43:48.241 --> 00:43:51.825
So at the moment something tips over, we're just like, cool, we'll just switch over to something else.

663
00:43:51.945 --> 00:43:53.947
So we have a great big blend of everything.

664
00:43:54.007 --> 00:43:57.070
I think we're a customer of just about all the labs at the moment, which is good fun.

665
00:43:57.963 --> 00:43:59.905
Fun and a nightmare at the same time, I'm sure.

666
00:44:00.006 --> 00:44:02.148
I'm sure that's got to be complicated to maintain.

667
00:44:02.168 --> 00:44:03.169
That's awesome.

668
00:44:03.530 --> 00:44:04.531
All right, man, this was awesome.

669
00:44:04.571 --> 00:44:06.373
Thanks so much for coming on.

670
00:44:07.735 --> 00:44:08.456
Crazy week.

671
00:44:08.536 --> 00:44:15.184
It's, like, good to talk to someone who's, like, literally on the front lines of AI security and code and everything that's, like, literally on fire right now.

672
00:44:15.304 --> 00:44:16.485
So...

673
00:44:17.313 --> 00:44:18.574
Yeah, it's fine.

674
00:44:18.834 --> 00:44:21.577
We'll chat again when we have another crazy week.

675
00:44:21.597 --> 00:44:22.017
For sure.

676
00:44:22.457 --> 00:44:24.639
When someone pulls the fire alarm again, we'll call you back.

677
00:44:25.000 --> 00:44:25.620
Exactly.

678
00:44:26.261 --> 00:44:27.061
All right, good time with you.

679
00:44:27.081 --> 00:44:27.662
We'll see you next time.

680
00:44:28.182 --> 00:44:28.442
See ya.

681
00:44:28.883 --> 00:44:32.406
Hey, hackers, before we keep going, I want to take a quick break and say a quick thank you to our sponsors.

682
00:44:32.666 --> 00:44:36.169
Today's Lowdown episode is sponsored by our friends over at Maze.

683
00:44:36.669 --> 00:44:41.953
Maze has spent the last couple of years using AI to find and remediate vulnerabilities in the cloud.

684
00:44:42.053 --> 00:44:48.818
And now they've released Maze Code, AI agents that find and deeply investigate vulnerabilities in your code.

685
00:44:49.038 --> 00:44:49.859
Now we all know the deal.

686
00:44:50.239 --> 00:44:56.484
Software composition analysis tools and SaaS toolings are all historically really bad at just making more noise.

687
00:44:56.524 --> 00:45:03.369
They find issues that may not actually be real and just create a lot of problems for your engineers to deal with that don't solve real security vulnerabilities.

688
00:45:03.469 --> 00:45:03.889
I get it.

689
00:45:03.929 --> 00:45:04.509
We've been there.

690
00:45:05.150 --> 00:45:10.514
Maze code is a really cool tool because Maze code actually knows how your code works.

691
00:45:10.854 --> 00:45:15.920
MazeCode investigates every finding with the context of how your code is actually deployed.

692
00:45:16.320 --> 00:45:23.909
Sure, you're using a version of libxml2 that's vulnerable, but the actual code path to hit that vulnerability is not exposed in your Docker image.

693
00:45:24.169 --> 00:45:27.673
MazeCode can find that and won't show you that that vulnerability is a big deal.

694
00:45:27.853 --> 00:45:32.415
Every vulnerability that Maze code reports comes with evidence that it's real and a way to fix it.

695
00:45:32.755 --> 00:45:37.097
Check out Maze at go.thelowdownpod.com slash Maze.

696
00:45:37.217 --> 00:45:40.899
That's go.thelowdownpod.com slash Maze.

697
00:45:41.079 --> 00:45:41.679
Thanks again, Maze.

698
00:45:41.820 --> 00:45:42.660
Let's keep going.

699
00:45:49.409 --> 00:45:51.011
Are we done with Black October?

700
00:45:51.031 --> 00:45:52.552
No, we're not done with Black October.

701
00:45:52.572 --> 00:45:58.378
There's all these other volumes that we didn't get to while we were in Vegas that is why we're calling it Black October.

702
00:45:58.418 --> 00:45:59.018
That was cool.

703
00:45:59.079 --> 00:45:59.659
Thanks, Harry.

704
00:45:59.679 --> 00:46:01.201
You're a legend.

705
00:46:01.221 --> 00:46:02.101
This was awesome.

706
00:46:03.163 --> 00:46:04.604
But Node to Shell.

707
00:46:04.664 --> 00:46:07.267
Okay, first of all, can we declare to Shell dead?

708
00:46:07.507 --> 00:46:08.127
Can we just stop?

709
00:46:08.167 --> 00:46:10.049
Can we just say the Node bug, please?

710
00:46:10.189 --> 00:46:11.150
Like, come on.

711
00:46:11.511 --> 00:46:12.171
They're all to Shell.

712
00:46:12.211 --> 00:46:12.772
It's RCE.

713
00:46:12.792 --> 00:46:13.272
Fuck you guys.

714
00:46:13.332 --> 00:46:13.673
Like, come on.

715
00:46:14.271 --> 00:46:22.700
I gave all sorts of detailed analysis of my feelings of everything going on this week across multiple social media channels, and my most viewed thing was your mom to shell.

716
00:46:22.920 --> 00:46:23.901
That's so funny.

717
00:46:23.961 --> 00:46:25.222
Yeah, so to shell.

718
00:46:26.244 --> 00:46:27.265
I blame Dirty Cow, bro.

719
00:46:27.285 --> 00:46:32.330
Dirty Cow set the precedent that only cool kids with their cool exploits get websites, and then ever since then it's been a shit show.

720
00:46:32.350 --> 00:46:32.470
What?

721
00:46:32.670 --> 00:46:33.670
Was it that?

722
00:46:33.711 --> 00:46:37.792
I mean, Heartbleed and Poodle and Shellshock were all before the... Yeah, you're right.

723
00:46:37.832 --> 00:46:41.094
Heartbleed was one of the first big branding campaigns for sure.

724
00:46:42.234 --> 00:46:43.635
When your vuln has a marketing team.

725
00:46:44.595 --> 00:46:44.936
Really?

726
00:46:45.596 --> 00:46:47.877
So, Node to Shell.

727
00:46:48.037 --> 00:46:52.019
And it's funny, the first headline I saw about this one was about Python.

728
00:46:52.919 --> 00:46:55.820
And I was like, bro, your Python's got JavaScript problems.

729
00:46:55.860 --> 00:46:56.461
What's going on?

730
00:46:56.501 --> 00:47:01.363
I heard you like JavaScript, so I put JavaScript in your Python so you could JavaScript while you Python...

731
00:47:01.583 --> 00:47:02.284
Yeah.

732
00:47:02.465 --> 00:47:07.693
So it seems like there's a package called, not broccoli, Brotley.

733
00:47:07.893 --> 00:47:08.293
Brotley.

734
00:47:08.334 --> 00:47:10.297
I've seen broccoli to shell, which is so funny.

735
00:47:10.317 --> 00:47:10.497
Yeah.

736
00:47:10.857 --> 00:47:16.787
And it seems to be one of those packages that's just squirreled everywhere and is in a whole lot of things.

737
00:47:17.467 --> 00:47:25.312
And as opposed to a lot of the other stuff that we're going to talk about this week, this one does have absolutely known exploitation going on right now.

738
00:47:25.372 --> 00:47:26.452
It's in the Sysikev.

739
00:47:27.373 --> 00:47:30.454
I'm seeing people I trust be like, I'm finding this on client sites.

740
00:47:31.195 --> 00:47:31.795
We got to fix this.

741
00:47:31.835 --> 00:47:32.396
What is this one?

742
00:47:33.376 --> 00:47:33.656
Yeah.

743
00:47:33.676 --> 00:47:37.518
So this is another issue of like a, again, I'm not going to use Log4j.

744
00:47:39.283 --> 00:47:47.391
in the way that people want me to use log for a day, but it's similar in the sense that, like, it is a supply chain issue in that you don't know if you have it in your code base potentially, right?

745
00:47:47.411 --> 00:47:57.960
The reason why it's Node to Shell, Python to Shell, Microsoft to Shell is Broccoli, not Broccoli, is a compression algorithm that is created by Google, right?

746
00:47:57.980 --> 00:47:58.641
There's a library.

747
00:47:58.661 --> 00:47:59.201
You can go look it up.

748
00:47:59.222 --> 00:48:00.242
It's open source on GitHub.

749
00:48:01.083 --> 00:48:03.404
And Node supports it, right?

750
00:48:03.464 --> 00:48:04.345
Python supports it.

751
00:48:04.365 --> 00:48:08.567
A lot of these web frameworks that are written on top of these frameworks support Brotley.

752
00:48:08.607 --> 00:48:09.927
And I'm going to say Brotley in accident.

753
00:48:09.947 --> 00:48:10.948
You're going to have to deal with it.

754
00:48:11.588 --> 00:48:12.048
Compression.

755
00:48:12.588 --> 00:48:15.110
The vulnerability is in the decompression, right?

756
00:48:15.150 --> 00:48:26.355
So if you are able to send a Brotley compressed malicious stream to a server that is vulnerable to this, the decompression will cause an out-of-bounds write and can cause RCE.

757
00:48:27.315 --> 00:48:34.660
The issue, I think, back to the PHP to shell is the end of the earth branding issue is similarly, this is not a bug that affects everything, right?

758
00:48:34.680 --> 00:48:40.624
Like I run a lot of applications that I maintain that are written either with node frontends or are written in Python.

759
00:48:40.644 --> 00:48:42.085
I'm a big Django guy.

760
00:48:42.105 --> 00:48:43.185
Fuck you guys if you hate that.

761
00:48:43.205 --> 00:48:43.786
Django's a shit.

762
00:48:45.807 --> 00:48:49.629
But Node out of the box does not support Brotley, right?

763
00:48:49.949 --> 00:48:51.110
But it's called Node to Shell.

764
00:48:51.190 --> 00:48:54.472
I was kind of having a panic attack when I was in Las Vegas trying to figure out if I was going to get exploited by this.

765
00:48:54.612 --> 00:48:55.732
I didn't.

766
00:48:56.172 --> 00:48:56.473
Yeah.

767
00:48:56.673 --> 00:48:58.073
So, again, it's real.

768
00:48:58.374 --> 00:49:08.379
The vulnerable cases for Node to Shell are if you overtly use the Brotley package in Node.js and consume a Brotley stream with this vulnerable version.

769
00:49:08.479 --> 00:49:09.139
That is bad.

770
00:49:09.920 --> 00:49:13.241
Or if you have a piece of code in Node.

771
00:49:13.501 --> 00:49:14.362
Say that four times fast.

772
00:49:14.842 --> 00:49:17.585
that will go out and run the fetch operation, right?

773
00:49:17.605 --> 00:49:29.498
The fetch operation will go out, and, sorry, a fetch operation where the URL is controlled by a user, right, by an attacker, because then the attacker can specify the fetch operation to go out and retrieve a stream that it will then ingest and exploit.

774
00:49:29.518 --> 00:49:30.919
So a couple of preconditions there.

775
00:49:30.939 --> 00:49:32.000
Same thing applies to Python, right?

776
00:49:32.020 --> 00:49:33.001
It's not out of the box completely.

777
00:49:33.021 --> 00:49:33.742
Um...

778
00:49:34.661 --> 00:49:39.309
But ultimately, it's another case of, like, name implies worse than actually is.

779
00:49:39.489 --> 00:49:45.199
And then even when it's exploitable, it's still a remote memory corruption, which is hard to fingerprint and exploit.

780
00:49:45.219 --> 00:49:47.944
So not good, not that bad.

781
00:49:48.336 --> 00:49:58.286
Another kind of interesting part of this one is it was found by OpenAI, like internal during like a red team exercise.

782
00:49:58.346 --> 00:49:59.628
They were the ones who found it.

783
00:50:00.409 --> 00:50:06.395
They published the initial disclosure on their patchtheplanet.org website about this, right?

784
00:50:06.655 --> 00:50:07.276
Oh, nice.

785
00:50:07.996 --> 00:50:16.279
So, which is kind of weird because now Sam Altman is coming out and saying, like, I don't think it's going to be possible to patch all the bugs.

786
00:50:16.419 --> 00:50:19.320
It's like, well, you have like a whole initiative about patching all the bugs, right?

787
00:50:19.360 --> 00:50:20.700
So, internally.

788
00:50:20.720 --> 00:50:22.261
It's kind of concerning that you're saying this right now.

789
00:50:23.001 --> 00:50:23.241
Yeah.

790
00:50:23.261 --> 00:50:25.902
Well, what's going on over there behind door number three?

791
00:50:25.922 --> 00:50:27.022
Yeah.

792
00:50:27.182 --> 00:50:28.303
So, yeah.

793
00:50:28.783 --> 00:50:28.963
Yeah.

794
00:50:29.784 --> 00:50:41.989
again, just like further evidence of there's some like gnarly vuln research going on at Frontier Model land when they're finding stuff like this that's like buried so deep into a lot of different things.

795
00:50:42.129 --> 00:50:44.030
But anyway, hopefully 2Shell dies.

796
00:50:44.090 --> 00:50:44.410
It won't.

797
00:50:44.590 --> 00:50:44.910
It won't.

798
00:50:44.930 --> 00:50:48.952
We're going to read so many more 2Shell things over the next couple of weeks.

799
00:50:49.172 --> 00:50:49.912
Code 2Shell.

800
00:50:50.752 --> 00:50:51.533
Yeah, just everything.

801
00:50:51.553 --> 00:50:51.913
Web 2Shell.

802
00:50:52.193 --> 00:50:53.674
Yeah, everything 2Shell.

803
00:50:54.154 --> 00:50:54.534
So...

804
00:50:55.715 --> 00:51:09.705
We talked about this one in Vegas where E-Mudra, Certificate Authority, during their ceremony, key generation ceremony, I wish I had some candles and crystals and then we could like meditate a little bit.

805
00:51:09.826 --> 00:51:12.708
Generate the key, make it bigger than the number four.

806
00:51:13.809 --> 00:51:23.055
Well, now we have a completely unrelated story, not even a week later, where another private key having to do with TLS...

807
00:51:24.096 --> 00:51:26.377
is the number four again.

808
00:51:26.577 --> 00:51:26.817
Yeah.

809
00:51:26.877 --> 00:51:36.640
So not even the same root cause, not even the same, like, oh yeah, they, they, it turns out now that we know about people can stub their toe this way, we found another way that someone subbed their toe.

810
00:51:36.700 --> 00:51:37.740
What the hell happened?

811
00:51:38.180 --> 00:51:47.903
I made a tweet where I was like, send me your wacky hacker name, first name, IP address, middle name, username, and last name, private key pen format, please.

812
00:51:48.244 --> 00:51:52.605
And every response I got was local host route for just fucking,

813
00:51:53.085 --> 00:51:56.427
Four has become the six seven of cybersecurity, which is so funny.

814
00:51:57.088 --> 00:51:59.730
Yeah, this is an interesting disclosure from Fox IO.

815
00:51:59.870 --> 00:52:03.672
I've been seeing a lot of their stuff come out about, like, just bugs and everything.

816
00:52:05.333 --> 00:52:08.756
So they're talking about this thing called Embed This Go Ahead.

817
00:52:08.916 --> 00:52:13.399
It's like a lightweight HTTP server a la BOA or Lighty.

818
00:52:15.360 --> 00:52:20.943
Now, the issue in the code base has to do with TLS, right, transport layer security, just like kind of the Emudra thing.

819
00:52:20.963 --> 00:52:29.347
So when I saw something, something, another provider, the number four, I was like, oh, truly this means, surely this means that this device just uses the Emudra CA.

820
00:52:29.467 --> 00:52:29.867
That's all it is.

821
00:52:29.907 --> 00:52:30.567
It's not a big deal.

822
00:52:30.587 --> 00:52:33.869
And, like, no, it is literally a different vulnerability entirely.

823
00:52:34.409 --> 00:52:44.474
So what's going on here is if you communicate to a go-ahead web server over TLS and establish a Diffie-Hellman key exchange over RSA,

824
00:52:45.014 --> 00:52:45.676
There is a, yes.

825
00:52:46.197 --> 00:52:46.958
Excuse me, low level.

826
00:52:47.219 --> 00:52:47.559
Yes.

827
00:52:47.660 --> 00:52:49.103
Question from the class.

828
00:52:49.383 --> 00:52:49.564
Yes.

829
00:52:49.844 --> 00:52:50.886
What's a go ahead web server?

830
00:52:50.906 --> 00:52:50.987
Yes.

831
00:52:51.640 --> 00:52:56.504
I just said, a go ahead web server is a, uh, in, it's a web server for embedded applications.

832
00:52:56.704 --> 00:52:57.885
Powers, powers.

833
00:52:57.945 --> 00:52:58.465
What is this?

834
00:52:58.805 --> 00:53:12.856
Um, millions of IOT firewall, OT and military devices, um, from Palo Alto, Comcast, D-Link, HP, Canon, Honeywell, GE, Oracle, Siemens, uh, Rockwell, DG, Lockheed Martin, Boeing, and more.

835
00:53:13.116 --> 00:53:15.357
So just those, just those, just a few.

836
00:53:15.998 --> 00:53:18.219
Um, so the issue is kind of hilarious.

837
00:53:18.239 --> 00:53:18.300
Um,

838
00:53:19.080 --> 00:53:23.903
When you create a TLS context, in this case, you're using embed TLS.

839
00:53:23.923 --> 00:53:26.725
It is a TLS library for embedded applications.

840
00:53:26.765 --> 00:53:30.307
It's much smaller, much more optimized for size than speed.

841
00:53:31.988 --> 00:53:46.958
When you establish a context doing Diffie-Hellman, you have to establish the prime, right, the agreed-upon prime that you're going to use to do the key exchange, and the G, the generator, right, to do the number generation so you can have your, you know, your private key and your public key to do the key exchange.

842
00:53:48.306 --> 00:54:00.679
When you use their API, when you establish this configuration, you have to say not only the variable, like the prime number, the pointer to that thing, it's specify the size of that thing.

843
00:54:00.699 --> 00:54:02.621
So you say this is a pointer to the prime.

844
00:54:02.902 --> 00:54:04.103
This is the size of the prime.

845
00:54:04.123 --> 00:54:05.604
This is a pointer to the G value.

846
00:54:05.644 --> 00:54:07.046
This is the size of the G value.

847
00:54:07.486 --> 00:54:09.609
The G value is only one byte.

848
00:54:10.029 --> 00:54:13.471
The generator, or sorry, the G value is only one byte.

849
00:54:13.511 --> 00:54:15.712
The prime number is 256 bytes.

850
00:54:15.752 --> 00:54:17.393
It gives you times eight, right?

851
00:54:17.433 --> 00:54:17.813
What is that?

852
00:54:17.853 --> 00:54:19.994
It's like 2048 bits of entropy.

853
00:54:20.214 --> 00:54:20.655
That's great.

854
00:54:21.492 --> 00:54:46.957
issue there is that when they put the size of for the Diffie Hellman prime, which should be two, five, six, they accidentally put in the size of the Diffie Hellman G, which is one, which gave the effective length of the prime of the prime number to be one byte, which means that the prime number was only ever the number four the entire time, which means you can take any connection,

855
00:54:52.218 --> 00:54:58.380
you know the prime so you can then extract the key information from the session, decrypt the session, and man in the middle all the traffic.

856
00:54:59.260 --> 00:54:59.601
Not great.

857
00:55:00.241 --> 00:55:00.681
Not very good.

858
00:55:01.381 --> 00:55:02.801
Famous prime number four.

859
00:55:03.262 --> 00:55:04.342
Yes, the famous prime number.

860
00:55:04.402 --> 00:55:04.822
Exactly.

861
00:55:05.382 --> 00:55:05.742
Exactly.

862
00:55:05.762 --> 00:55:10.364
So it's not only a bad Diffie Hellman prime, it's

863
00:55:11.466 --> 00:55:11.846
Not a prime.

864
00:55:12.347 --> 00:55:14.568
So not exactly the strongest Tiffy Hellman.

865
00:55:14.869 --> 00:55:23.635
So this is not in a certificate of authority like last week where like, oh, browsers had to rush to pass all of a sudden the wifi pineapple attack gets more interesting.

866
00:55:23.815 --> 00:55:27.198
All of a sudden, like you said, like BGP type interception gets more interesting.

867
00:55:27.538 --> 00:55:30.700
What's the actual attack scenario for this one?

868
00:55:30.740 --> 00:55:36.985
Obviously we listed a bunch of scary company names that, that this kind of stuff could be found in IOT devices.

869
00:55:37.485 --> 00:55:40.808
So this is just, if you have man in the middle on any network, uh,

870
00:55:41.828 --> 00:55:45.830
and then traffic is getting negotiated to one of these devices?

871
00:55:46.110 --> 00:55:47.310
Is that the attack?

872
00:55:48.131 --> 00:55:49.851
You're just wire-sharking any network?

873
00:55:49.871 --> 00:55:52.812
You can decrypt any traffic that goes to these things specifically?

874
00:55:53.052 --> 00:56:05.797
Yeah, because of the broken Diffie-Hellman parameter, it's in the article, but basically the established shared secret over Diffie-Hellman when you have a prime of four, which isn't a prime, it reduces the shared secrets down to these numbers, 1, 2, 4, 8, 10, 20, 40, 80 in hex.

875
00:56:08.270 --> 00:56:13.733
So you have a potential eight different key values that can be used over this Diffie Hellman, which is not good.

876
00:56:13.793 --> 00:56:21.697
So, yeah, anyone sitting in the middle that can capture encrypted, quote-unquote, traffic between a client and a go-ahead router can be decrypted.

877
00:56:21.977 --> 00:56:23.058
And then, yeah, man in the middle, too.

878
00:56:23.098 --> 00:56:29.842
Like, you can arbitrarily insert yourself as a trusted authority, right, over an exchange and do whatever malicious stuff you want to do.

879
00:56:30.082 --> 00:56:31.503
Bad, bad, bad, bad, bad, bad, bad.

880
00:56:31.803 --> 00:56:33.664
I just literally couldn't believe that it was for again.

881
00:56:33.684 --> 00:56:33.944
Yeah.

882
00:56:34.562 --> 00:56:37.545
The fact that it's four, again, I thought it was related to Imudra.

883
00:56:37.565 --> 00:56:39.867
Like, oh, that means that they're still trusting that CA.

884
00:56:39.927 --> 00:56:41.809
No, it is literally a different bug.

885
00:56:42.129 --> 00:56:42.710
Do you listen to that?

886
00:56:43.010 --> 00:56:43.991
You ever watch Bo Burnham?

887
00:56:44.472 --> 00:56:44.712
Yes.

888
00:56:45.032 --> 00:56:53.740
Okay, so when I saw a friend of the show, Ice Solstice, you know, Ice Solstice on Twitter, he was like, what the fuck?

889
00:56:53.780 --> 00:56:54.621
It's four again.

890
00:56:54.781 --> 00:56:58.305
I pictured that Bo Burnham song about country music.

891
00:56:58.765 --> 00:56:59.526
Dirt road.

892
00:56:59.566 --> 00:56:59.846
Yeah.

893
00:56:59.886 --> 00:57:00.146
Yeah.

894
00:57:00.426 --> 00:57:02.547
And when he goes, Oh, that's a scarecrow.

895
00:57:02.587 --> 00:57:06.529
And then later in the song, he goes, Oh, it's a scarecrow again.

896
00:57:06.589 --> 00:57:06.870
Again.

897
00:57:07.270 --> 00:57:10.712
I, it's one of the hardest I've ever laughed at any comedy special ever again.

898
00:57:10.832 --> 00:57:11.052
Right.

899
00:57:11.212 --> 00:57:15.154
And, and I've been trying to think of like how to make the meme, like it's the scarecrow again.

900
00:57:15.214 --> 00:57:16.315
And it'd be the number four.

901
00:57:16.415 --> 00:57:19.436
And I haven't quite figured it out yet, but you know what?

902
00:57:19.496 --> 00:57:21.698
I know a funny, a personal fun fact about me.

903
00:57:22.433 --> 00:57:25.175
You know the Netflix special where he shows that song?

904
00:57:25.635 --> 00:57:25.836
Yeah.

905
00:57:26.196 --> 00:57:27.076
I'm in that special.

906
00:57:27.136 --> 00:57:27.837
No.

907
00:57:27.917 --> 00:57:28.778
Yeah, dude.

908
00:57:28.798 --> 00:57:30.759
When he filmed it in Port Chester, New York, I was there.

909
00:57:31.320 --> 00:57:32.120
Jelly.

910
00:57:32.160 --> 00:57:32.500
Yeah, dude.

911
00:57:32.560 --> 00:57:33.041
It was so good.

912
00:57:33.701 --> 00:57:36.303
That is the hardest that I've laughed in my entire life up to that point.

913
00:57:36.323 --> 00:57:36.583
So good.

914
00:57:36.603 --> 00:57:37.304
Oh, my God.

915
00:57:37.344 --> 00:57:37.744
So good.

916
00:57:38.165 --> 00:57:41.727
His special Inside during COVID was on repeat in my house.

917
00:57:41.767 --> 00:57:44.930
To just give you a little glimpse into my mental state at that point.

918
00:57:45.130 --> 00:57:46.671
It was great, but it was rough.

919
00:57:47.251 --> 00:57:48.172
It was so rough.

920
00:57:48.552 --> 00:57:49.812
If you haven't watched it, go watch it.

921
00:57:50.773 --> 00:57:53.954
And then be like, Matt, watch this on repeat for weeks.

922
00:57:54.094 --> 00:57:54.674
Yes, I did.

923
00:57:55.094 --> 00:57:56.395
Yes, on repeat for weeks.

924
00:57:57.475 --> 00:57:57.755
Okay.

925
00:57:57.975 --> 00:58:05.218
So can't believe we're just using the famous prime number four multiple times with the ability to decrypt traffic.

926
00:58:05.398 --> 00:58:07.039
It's a fucking scarecrow again.

927
00:58:07.079 --> 00:58:08.740
It's a fucking scarecrow again.

928
00:58:10.482 --> 00:58:11.402
One of my favorite jokes.

929
00:58:11.962 --> 00:58:12.603
Oh, shit.

930
00:58:12.703 --> 00:58:30.989
Okay, so we have some other news from this week, but I did want and we did promise our loyal listeners who obviously were taking notes during our Vegas episode and they would otherwise burn us at the stake if we didn't go cover some of the stuff that we just did not have time to during the famous Black October week.

931
00:58:31.289 --> 00:58:31.969
It would go burn us.

932
00:58:32.592 --> 00:58:36.635
They would Bo Burnham, Bo Burnham, Burnham.

933
00:58:37.076 --> 00:58:45.222
So there were two other breaches that happened like right around PHP time that I think are even on the Wikipedia page for Black October.

934
00:58:45.848 --> 00:58:46.128
Maybe?

935
00:58:46.248 --> 00:58:47.169
Yeah, LightLM.

936
00:58:47.409 --> 00:58:48.569
Oh, no, Bratley was one.

937
00:58:48.589 --> 00:58:49.309
GitLab was one.

938
00:58:49.990 --> 00:58:50.470
Emudra.

939
00:58:50.490 --> 00:58:51.470
So we talked about all those.

940
00:58:51.550 --> 00:58:53.751
We did not get to talk about these two.

941
00:58:53.831 --> 00:58:59.154
So number one, you mentioned it earlier in the show, OpenCode got hacked.

942
00:58:59.774 --> 00:59:01.394
I made a standalone video on this.

943
00:59:01.414 --> 00:59:01.735
Did you?

944
00:59:01.875 --> 00:59:02.875
I forget if you covered this one.

945
00:59:02.895 --> 00:59:03.535
I know Prime did.

946
00:59:03.555 --> 00:59:04.776
No, I didn't cover it for two reasons.

947
00:59:04.876 --> 00:59:05.916
One, I have no details.

948
00:59:06.056 --> 00:59:07.037
I don't know what happened.

949
00:59:07.077 --> 00:59:08.137
And do I have friends there?

950
00:59:08.157 --> 00:59:10.658
And I don't want to add more media if they don't want it.

951
00:59:10.678 --> 00:59:11.599
And I haven't talked to them yet.

952
00:59:11.619 --> 00:59:11.719
Sure.

953
00:59:11.799 --> 00:59:11.899
Okay.

954
00:59:11.919 --> 00:59:12.339
So yeah.

955
00:59:12.659 --> 00:59:13.519
Sorry, I'm about to.

956
00:59:13.800 --> 00:59:14.240
No, no, no.

957
00:59:14.280 --> 00:59:16.281
I had to make a separate video, but talking about it here, I don't care.

958
00:59:16.301 --> 00:59:19.902
I'm not going to make a fucking headline like, open code situation is crazy.

959
00:59:19.922 --> 00:59:20.663
I'm not doing that.

960
00:59:21.623 --> 00:59:24.384
Sorry, I did.

961
00:59:24.925 --> 00:59:27.126
And the guy that you're about to go record with did too.

962
00:59:27.186 --> 00:59:30.547
Prime made a really good video about this.

963
00:59:31.608 --> 00:59:33.669
And I'm in the Prime video.

964
00:59:33.709 --> 00:59:34.709
Did you see my tweet about this?

965
00:59:34.889 --> 00:59:35.149
I did.

966
00:59:35.169 --> 00:59:36.310
Yeah, I saw your jump scare tweet.

967
00:59:36.470 --> 00:59:42.115
Dude, I was watching the Prime video just because I do watch his channel all the time and he's really good at doing the thing that I try to do.

968
00:59:42.135 --> 00:59:44.177
And I'm like, all right, how did he cover?

969
00:59:44.557 --> 00:59:45.998
He's covering hacks again.

970
00:59:46.018 --> 00:59:50.542
This fucking developer who's really good at YouTube is like doing cyber better than me.

971
00:59:50.642 --> 00:59:51.944
Crap, what's going on, right?

972
00:59:52.504 --> 00:59:54.545
And then all of a sudden, my face is on it.

973
00:59:54.565 --> 00:59:56.145
He screenshotted my tweet.

974
00:59:56.245 --> 00:59:58.166
Okay, so yeah, agree.

975
00:59:58.506 --> 01:00:00.766
I didn't make a deep dive video on this one for that same reason.

976
01:00:00.806 --> 01:00:02.047
It's like, what happened?

977
01:00:02.107 --> 01:00:02.767
We do not know.

978
01:00:02.827 --> 01:00:06.588
We still do not know what the root cause of this breach was.

979
01:00:06.888 --> 01:00:09.149
Zooming out, what's open code, everybody?

980
01:00:10.149 --> 01:00:11.709
It's a way that it's like a harness, right?

981
01:00:11.729 --> 01:00:12.610
Is that fair to call it that?

982
01:00:12.710 --> 01:00:15.910
Of a way that people use a lot of these local models.

983
01:00:16.090 --> 01:00:17.271
Is it similar to Olama?

984
01:00:17.451 --> 01:00:17.931
Would you like...

985
01:00:18.251 --> 01:00:22.393
No, I would say OpenCode is like a not-anthropic version of CloudCode.

986
01:00:22.473 --> 01:00:23.014
Okay, yeah.

987
01:00:23.034 --> 01:00:23.374
Same vibe.

988
01:00:23.614 --> 01:00:24.814
Got it, got it, got it, got it, got it.

989
01:00:24.834 --> 01:00:26.155
But they also do a lot of hosting, right?

990
01:00:26.175 --> 01:00:34.940
So like they not only have their own TUI where you can do CloudCode stuff, but they do a lot of self-hosting, not self-hosting, but like cloud hosting of models for you.

991
01:00:34.960 --> 01:00:38.582
You can use DeepSeek and GLM, I think, for free right now on OpenCode.

992
01:00:38.622 --> 01:00:39.442
Not an ad, but hit us up.

993
01:00:39.502 --> 01:00:40.662
That's what I was about to say, right?

994
01:00:40.722 --> 01:00:47.705
So, like, yeah, you can use it to do these local models, like famously local models, open weight models, like the GLM stuff and the DeepSeek stuff.

995
01:00:47.845 --> 01:00:58.868
But they are hosting them, and they caught so much attention the week before their hack because they were giving people free access to these good models that are, by the way, very good at cyber stuff, too, right?

996
01:00:58.868 --> 01:00:59.188
right?

997
01:00:59.208 --> 01:01:00.530
GLM 5.3, you've talked a lot about it.

998
01:01:00.930 --> 01:01:03.533
Deep Seek 4.1 Flash, really good at cyber stuff.

999
01:01:04.534 --> 01:01:13.443
And so, they caught, like, tons of attention and probably their user base went hockey stick over their, you know, ability to use this stuff for free.

1000
01:01:13.944 --> 01:01:15.305
Dot, dot, dot, two days later,

1001
01:01:16.046 --> 01:01:17.447
compromise headline, right?

1002
01:01:18.787 --> 01:01:31.492
The compromise was blown a bit out of proportion for like the, like I think they said it was like 1% of their user base because it wasn't like, oh, you use open code on your device and like you were using a bunch of stuff, like you're screwed.

1003
01:01:31.612 --> 01:01:38.135
No, they do have like a console though, like in their kind of web cloud version where you can upload your

1004
01:01:39.935 --> 01:01:45.218
or tie your API keys to the other frontier models and use OpenCode with the frontier models.

1005
01:01:45.778 --> 01:01:47.279
And you could like link it via this console.

1006
01:01:47.759 --> 01:01:49.800
So that was what was hacked.

1007
01:01:49.981 --> 01:02:03.828
Anyone who used the console and tied their frontier models, API keys to it, the things that was stolen was users of OpenCode's API keys to OpenAI or Anthropic were what was targeted during the attack.

1008
01:02:04.008 --> 01:02:04.308
Okay.

1009
01:02:05.035 --> 01:02:11.801
So the, and the only detail we know was that it was in a legacy on its way to be depreciated system.

1010
01:02:12.642 --> 01:02:13.583
We don't know what that means.

1011
01:02:14.203 --> 01:02:19.508
Like it doesn't matter because obviously production API keys were still in it kicking around for customers.

1012
01:02:20.089 --> 01:02:22.851
So legacy or not, that's what happened.

1013
01:02:22.971 --> 01:02:25.513
We do not like, they're like ongoing investigation.

1014
01:02:25.553 --> 01:02:26.434
We can't say anymore.

1015
01:02:26.735 --> 01:02:27.015
Okay.

1016
01:02:27.435 --> 01:02:27.655
Right.

1017
01:02:28.316 --> 01:02:31.499
I hope that they do put out some more transparency around like what actually went down.

1018
01:02:32.740 --> 01:02:38.986
Then they started their announcement on it was that this was due to a distillation attack.

1019
01:02:39.246 --> 01:02:47.655
And this is the tweet that I made that wound up in the prime video was, how does a distillation attack on open code lead to API keys of your customers getting stolen?

1020
01:02:48.035 --> 01:02:54.982
If you guys are unfamiliar, distillation attacks are what frontier models are undergoing from a lot of the Chinese models.

1021
01:02:55.743 --> 01:03:06.745
Anthropic and OpenAI are basically pointing fingers at a lot of these open-weight Chinese models saying that these are the results of distilling the Frontier models.

1022
01:03:07.605 --> 01:03:07.985
Yeah.

1023
01:03:08.645 --> 01:03:21.748
So I'm not like an expert on this, but basically they're like, they have a series of prompts that they use to train like a knockoff version of the Frontier model based on the responses to a lot of these prompts.

1024
01:03:21.768 --> 01:03:21.908
Yeah.

1025
01:03:22.388 --> 01:03:26.792
Yeah, the ultimate, I think, goal is to like, you're literally trying to extract the weights of the model.

1026
01:03:27.112 --> 01:03:28.373
How you do that, I have no idea.

1027
01:03:28.393 --> 01:03:29.333
Right, right, right.

1028
01:03:29.374 --> 01:03:30.995
Yeah, that's the part I'm not an expert in.

1029
01:03:31.015 --> 01:03:41.022
But like, apparently, and so the announcement by OpenCode was written a little weird that a bunch of us thought, wait, how did you guys get attacked with a distillation attack?

1030
01:03:41.122 --> 01:03:43.924
And then that led to customer API key losses.

1031
01:03:45.566 --> 01:03:51.330
Other way around, and it was just worded weird, was, and thank you for everyone who corrected me on that one.

1032
01:03:52.735 --> 01:04:02.117
the end goal of stealing these customers' API keys was to use them as part of a broader distillation attack against OpenAI and Anthropic.

1033
01:04:02.317 --> 01:04:02.557
Yeah.

1034
01:04:02.757 --> 01:04:17.261
So it turns out because the amount of prompts that you have to hit the frontier model with in order to try to extract the models, the weights, sorry, is so high, you're basically spamming the frontier models and you're going to get your API key banned.

1035
01:04:17.361 --> 01:04:18.561
It's obvious what you're doing.

1036
01:04:18.921 --> 01:04:21.242
So in order to pull off these distillation attacks,

1037
01:04:22.602 --> 01:04:29.927
a swath of legitimate keys to like use as part of the attack before they get blocked as part of the attack.

1038
01:04:30.187 --> 01:04:32.709
And so that's actually how OpenCode found out they got hacked.

1039
01:04:32.769 --> 01:04:41.694
OpenAI started notifying people's tokens like, hey, like notifying people's accounts like, hey, your API key is being used in a distillation attack and we're shutting it down.

1040
01:04:41.714 --> 01:04:43.035
And people are like, what the hell is going on?

1041
01:04:43.456 --> 01:04:43.896
Interesting.

1042
01:04:44.378 --> 01:04:50.322
It's almost like distillation accounts for the new crypto miners in that when you compromise, what happens, right?

1043
01:04:50.342 --> 01:04:54.884
If they're not trying to steal your stuff or hold you at ransom, they're just going to use you to mine Ethereum or Solana or whatever.

1044
01:04:55.405 --> 01:05:00.367
It's almost like now the new crypto miner hotness is like, do you have an API key that I can distill with?

1045
01:05:00.488 --> 01:05:00.808
Hmm.

1046
01:05:01.368 --> 01:05:03.270
You know, the new black market.

1047
01:05:03.430 --> 01:05:07.694
Yeah, way to slightly monetize, you know, some InfoStealer results, right?

1048
01:05:08.354 --> 01:05:13.459
We're even seeing InfoStealers update and target these API keys because they're valuable, right?

1049
01:05:14.139 --> 01:05:17.182
Either in use in attacks or use in distillation attacks.

1050
01:05:17.983 --> 01:05:19.504
So anyway, we didn't cover this one.

1051
01:05:19.644 --> 01:05:20.425
Wanted to cover this one.

1052
01:05:20.585 --> 01:05:21.306
Kind of interesting.

1053
01:05:22.267 --> 01:05:25.870
I'm hoping we hear more about like how it actually went down, but whatever.

1054
01:05:25.890 --> 01:05:26.050
Yeah.

1055
01:05:26.727 --> 01:05:33.193
The other big one that we didn't get to cover in Vegas was Light LLM had another supply chain attack.

1056
01:05:33.213 --> 01:05:35.095
So this happened already in the spring.

1057
01:05:35.596 --> 01:05:38.319
We probably covered it on like an early podcast.

1058
01:05:38.339 --> 01:05:39.600
Maybe it was before we started the podcast.

1059
01:05:39.920 --> 01:05:41.542
I know I made videos about the original.

1060
01:05:42.062 --> 01:05:42.642
version of this.

1061
01:05:43.143 --> 01:05:48.846
But if you guys don't know, LightLM is like a gateway to a lot of these things, a lot of these models.

1062
01:05:48.946 --> 01:05:50.647
It's highly, highly popular.

1063
01:05:51.047 --> 01:05:56.590
It's like tens of thousands of GitHub stars, 92 million monthly downloads on PyPy, right?

1064
01:05:56.710 --> 01:06:01.913
So this is like a Python SDK, an AI gateway.

1065
01:06:02.693 --> 01:06:09.138
And twice now it's, it's had a supply chain attack where it was started to push malware.

1066
01:06:09.439 --> 01:06:12.701
We saw like the big team PCP wave of supply chain stuff.

1067
01:06:12.781 --> 01:06:13.882
It started to quiet down.

1068
01:06:14.322 --> 01:06:20.087
Nothing architecturally has changed about PyPy or NPM or the, or the supply chain world in general.

1069
01:06:20.107 --> 01:06:24.530
So these attacks are still very valid and easy to pull off.

1070
01:06:24.950 --> 01:06:25.151
Right.

1071
01:06:25.211 --> 01:06:27.152
It's just team PCP got arrested.

1072
01:06:27.252 --> 01:06:29.034
So it's gotten quieter.

1073
01:06:29.054 --> 01:06:29.214
Yeah.

1074
01:06:29.754 --> 01:06:30.875
But that's what happened.

1075
01:06:30.915 --> 01:06:37.903
Someone compromised Lite LLMs, either their build pipeline or their GitHub accounts or whatever it was.

1076
01:06:37.923 --> 01:06:44.469
And instead of worming, like we saw all year, it was the supply chain worms.

1077
01:06:44.690 --> 01:06:45.931
Yeah, the Shia Lude stuff.

1078
01:06:45.951 --> 01:06:51.677
Instead of worming, they just used the official Lite LLM releases to distribute...

1079
01:06:52.177 --> 01:06:52.657
Here we go.

1080
01:06:52.917 --> 01:06:54.278
Info Stealer malware, right?

1081
01:06:54.718 --> 01:07:01.901
And the Info Stealer malware was targeting cloud deployment credentials and AI credentials and then like developer secrets type stuff.

1082
01:07:02.201 --> 01:07:02.841
There you go.

1083
01:07:02.881 --> 01:07:03.642
Were like the main things.

1084
01:07:03.682 --> 01:07:07.483
And then normal Info Stealer stuff, browser data, cookies, host information, whatever.

1085
01:07:08.683 --> 01:07:12.025
The interesting part about all of these supply chain attacks that we've seen

1086
01:07:13.145 --> 01:07:17.989
And we talked about this in our did Team PCP do anything wrong segment, right?

1087
01:07:18.329 --> 01:07:21.111
Was we haven't seen a lot of like ripple effects, right?

1088
01:07:21.131 --> 01:07:23.252
So it's like 92 million monthly downloads.

1089
01:07:23.473 --> 01:07:27.115
Obviously, you know, people shut this down pretty quick.

1090
01:07:27.796 --> 01:07:33.640
But some number of that 92 million probably sucked up this malware as part of the official LightLM release.

1091
01:07:34.100 --> 01:07:38.263
And it got their cloud credentials and AI credentials and dev secrets and stuff stolen.

1092
01:07:38.964 --> 01:07:39.344
Now what?

1093
01:07:39.684 --> 01:07:45.168
It's not like those people all instantly rotated every secret and then all the secrets that were stolen during this were not valuable.

1094
01:07:45.528 --> 01:07:52.073
So there should be a bunch of ripple second order effects of the threat actor using the stuff that they stole out of this.

1095
01:07:52.373 --> 01:07:53.354
And we're not really seeing that.

1096
01:07:53.374 --> 01:07:54.115
We don't really see it.

1097
01:07:55.295 --> 01:07:58.598
And with Team PCP, they came out and said, well, that was never their goal.

1098
01:07:58.618 --> 01:07:59.919
Their goal was to kind of...

1099
01:08:00.786 --> 01:08:04.388
be the Robin Hood hacker and point out that the house of cards is going on here.

1100
01:08:04.408 --> 01:08:07.870
Debatable, right?

1101
01:08:08.030 --> 01:08:08.810
But whatever.

1102
01:08:10.371 --> 01:08:20.536
Ironically, Railway also got hacked, which is confusing because the Light LLM logo is a train, and I was like, wait, does Railway write Light LLM?

1103
01:08:20.556 --> 01:08:21.596
Did they get hacked because of that?

1104
01:08:21.876 --> 01:08:22.117
Nope.

1105
01:08:22.137 --> 01:08:25.318
Completely separate events that are just muddled by the same logo.

1106
01:08:25.338 --> 01:08:26.679
So, different thing for sure.

1107
01:08:27.599 --> 01:08:29.541
The railway hack was kind of nuts, too, by the way.

1108
01:08:30.042 --> 01:08:31.143
We don't have it in our show notes.

1109
01:08:31.283 --> 01:08:32.704
Yeah, I have no idea how they got in.

1110
01:08:32.744 --> 01:08:34.866
Bad, bad, bad, bad, bad, bad, bad, bad.

1111
01:08:35.267 --> 01:08:36.748
Yeah, man, supply chain go brr.

1112
01:08:36.828 --> 01:08:42.814
I don't know what to say about these anymore because it's just like, all right, another AI package got compromised.

1113
01:08:42.934 --> 01:08:43.574
God damn.

1114
01:08:44.035 --> 01:08:46.137
Make sure you're updating your packages, I guess.

1115
01:08:46.565 --> 01:08:46.845
Yeah.

1116
01:08:47.145 --> 01:08:50.788
I guess one of my takeaways, I know I kind of lumped Light Elim and OpenCode.

1117
01:08:50.808 --> 01:08:55.512
They're not really related, but one of my takeaways, especially from the OpenCode one, I know you have friends that work there.

1118
01:08:55.532 --> 01:08:57.754
I know Dax is big on Twitter.

1119
01:08:57.934 --> 01:09:03.378
Seems like a great dude, developer, member of the community, not dunking on anything.

1120
01:09:03.538 --> 01:09:09.823
Dax's response to this was like, okay, got to do a bunch of cybersecurity hardening now, right?

1121
01:09:10.043 --> 01:09:10.524
And it's like,

1122
01:09:11.786 --> 01:09:13.407
Yeah, that's like day one.

1123
01:09:13.427 --> 01:09:14.727
You have to do that now, right?

1124
01:09:14.787 --> 01:09:18.788
Is like the lesson I'm taking away from a lot of this is like the speed of this.

1125
01:09:19.028 --> 01:09:27.952
The other thing is like a lot of people were using this for pretty sensitive stuff with their like prod like AI keys and they were losing thousands of dollars during this attack.

1126
01:09:28.452 --> 01:09:34.914
And it's like you people realize like you're using a tool that's like put out and you're tying sensitive API keys to it.

1127
01:09:35.884 --> 01:09:38.346
Like that company does not have a security team, right?

1128
01:09:38.386 --> 01:09:44.970
Like these are just kind of viral products put out by engineers that are like have a community and an audience and all this kind of stuff.

1129
01:09:44.990 --> 01:09:45.490
And I get it.

1130
01:09:45.510 --> 01:09:51.433
I'm not like discouraging using things like this or experimenting or anything like that.

1131
01:09:51.453 --> 01:09:58.417
It's just like we have to know what we're getting into here and what our threat model is when we're plugging in a lot of like prod stuff and API keys into...

1132
01:09:59.478 --> 01:10:01.861
pretty fly-by-the-seat-of-your-pants type projects, right?

1133
01:10:01.921 --> 01:10:05.165
It is interesting that people are using their prod API key as an open code.

1134
01:10:05.306 --> 01:10:10.292
Just like, in general, not even from a security standpoint, just like, you know, separation of responsibility, right?

1135
01:10:10.332 --> 01:10:14.157
Like, why is your prod key being used for dev stuff?

1136
01:10:14.197 --> 01:10:14.758
It's a little weird.

1137
01:10:15.158 --> 01:10:15.378
Yeah.

1138
01:10:15.618 --> 01:10:21.001
And so like, I mean, people were waking up to thousand, you know, $1,700 bills, you know, multi-thousand dollar bills.

1139
01:10:21.021 --> 01:10:23.362
I got DMs, people being like, I was part of this, you know.

1140
01:10:24.923 --> 01:10:27.324
So yeah, it's kind of wild.

1141
01:10:27.364 --> 01:10:29.726
And then the LightLM thing, same thing.

1142
01:10:29.766 --> 01:10:33.488
It's like we're using these projects that are like popular on GitHub as part of our stamp.

1143
01:10:33.568 --> 01:10:40.732
And then I think the guidance on a lot of these type of things, these supply chain attacks, is to limit your exposure.

1144
01:10:40.812 --> 01:10:43.153
Don't pull the update day one.

1145
01:10:43.173 --> 01:10:43.553
Right.

1146
01:10:44.234 --> 01:10:53.502
give it like, I think five to seven days right now is like the agreed upon, okay, we'll probably catch the malware in like five to seven days.

1147
01:10:54.743 --> 01:10:59.146
And you won't have, because vulnerabilities get patched in these updates.

1148
01:10:59.166 --> 01:11:02.950
So you don't want to wait too long to pull the new version of something because you're going to have CVEs.

1149
01:11:02.970 --> 01:11:04.631
If you pull too fast, you might get malware.

1150
01:11:04.991 --> 01:11:07.033
If you pull too slow, you're going to have CVEs for too long.

1151
01:11:07.414 --> 01:11:11.277
So there's like this few day window where everyone's kind of like, yeah, that's the sweet spot right now.

1152
01:11:11.397 --> 01:11:12.377
Yeah, it's a balancing act, right?

1153
01:11:12.397 --> 01:11:18.599
It's like if there's a 10.0 critical CVSS RCE in a library that you use user-facing, all right, maybe we update that one.

1154
01:11:18.959 --> 01:11:26.302
But if it's like, oh, like minor perf update, like, okay, I think five to 14 days is probably a reasonable wait.

1155
01:11:27.002 --> 01:11:31.025
I'll tell you, for the most part, a lot of people aren't choosing per update, right?

1156
01:11:31.085 --> 01:11:32.567
This stuff is all just managed, right?

1157
01:11:32.627 --> 01:11:33.027
Oh, for sure.

1158
01:11:33.167 --> 01:11:35.849
Package.json pinned to latest, right?

1159
01:11:35.929 --> 01:11:46.898
Like, or whatever it is, you know, that's why I think, I mean, there's some cool tools in this space, but I think the pull after five or seven days is probably a pretty good best practice at this point.

1160
01:11:47.238 --> 01:11:51.740
Yeah, but like I'm saying, though, if an RCE comes out, you can just go in there and manually pull it up.

1161
01:11:52.540 --> 01:11:56.041
The vast majority of people are not doing that, as per 32 million vulnerable PHP.

1162
01:11:56.061 --> 01:11:57.882
As per 32 million PHP to shelves.

1163
01:11:58.442 --> 01:11:59.402
All right, man.

1164
01:11:59.543 --> 01:12:03.244
You made a video about this Broadcom thing that I was super interested in.

1165
01:12:03.384 --> 01:12:07.345
It was like the Wi-Fi router type stuff going on.

1166
01:12:07.405 --> 01:12:12.727
And then there's this other one, UEFI firmware stuff.

1167
01:12:13.027 --> 01:12:14.048
This is obviously...

1168
01:12:14.728 --> 01:12:19.171
a little low level for me on the hardware stuff.

1169
01:12:19.191 --> 01:12:21.112
This is more your area of expertise.

1170
01:12:21.212 --> 01:12:21.693
What was going on?

1171
01:12:21.713 --> 01:12:24.574
We didn't want to have Broadcom in the show notes, but I just remembered that you made that video.

1172
01:12:24.594 --> 01:12:27.817
I don't know if you want to lump it in with this, but what's going on in the hardware one?

1173
01:12:27.837 --> 01:12:29.458
I'm trying to remember what the bug for Broadcom even is.

1174
01:12:29.518 --> 01:12:30.058
I can talk about UEFI.

1175
01:12:30.078 --> 01:12:32.319
Broadcom is like, that's so five days ago.

1176
01:12:32.339 --> 01:12:32.900
Like, I don't know.

1177
01:12:33.840 --> 01:12:34.961
Black October, Ed.

1178
01:12:35.001 --> 01:12:35.401
Black October.

1179
01:12:35.421 --> 01:12:35.781
I know.

1180
01:12:35.861 --> 01:12:36.582
How dare I?

1181
01:12:36.762 --> 01:12:38.163
So the Broadcom bug is pretty simple.

1182
01:12:38.183 --> 01:12:46.667
It was just like in the code for the Broadcom chipset, the code that handled the Wi-Fi protected setup like controller had a buffer overflow.

1183
01:12:47.067 --> 01:12:47.687
Not a big deal.

1184
01:12:48.088 --> 01:13:00.594
The thing that was kind of spicy about it was that Broadcom like knew about it for coming up on I think seven years and like just didn't issue a CVE because like, oh, like that component is deprecated and like we don't support it anymore.

1185
01:13:00.634 --> 01:13:01.935
So we don't have to issue a CVE.

1186
01:13:01.955 --> 01:13:02.175
Yeah.

1187
01:13:02.675 --> 01:13:09.416
It's like a, the entire plumbing of the cybersecurity ecosystem rides on you guys publishing CBEs.

1188
01:13:09.897 --> 01:13:13.977
And two, you are a manufacturer of wireless chipsets.

1189
01:13:14.457 --> 01:13:19.338
You know, for a fact that once the firmware goes into those things, they're, they're likely not catching updates.

1190
01:13:19.698 --> 01:13:22.219
So a lot of the world is still using, what is it?

1191
01:13:22.519 --> 01:13:25.600
WPS monitor D I think is the binary in question.

1192
01:13:25.620 --> 01:13:26.460
Um, yeah.

1193
01:13:27.220 --> 01:13:32.401
So it affects a lot of people, and the problem is you don't even have to have WPS on or enabled.

1194
01:13:32.421 --> 01:13:36.882
This lives in the system on a chip before the OS stack gets to it.

1195
01:13:36.922 --> 01:13:45.524
So if you can basically drive up in an unmarked FBI van to someone's house where this router exists and has this bug, you can get code execution privileged on the router.

1196
01:13:45.644 --> 01:13:46.785
Not great.

1197
01:13:46.805 --> 01:13:55.527
A different bug than the UEFI one that another nickel in the overhyped sensationalist title bucket—

1198
01:13:56.507 --> 01:14:01.430
It's not good that we have a code execution vulnerability in UEFI firmware.

1199
01:14:01.850 --> 01:14:02.571
Obviously, right?

1200
01:14:02.891 --> 01:14:03.331
Not great.

1201
01:14:03.812 --> 01:14:06.693
But the problem here is kind of the way we said it.

1202
01:14:06.773 --> 01:14:09.375
It affects most servers on planet Earth.

1203
01:14:09.915 --> 01:14:10.156
Sure.

1204
01:14:10.636 --> 01:14:18.881
So this is a vulnerability in the UEFI firmware, which to remind everybody, like when your computer turns on, it turns on at the most privileged level.

1205
01:14:18.941 --> 01:14:23.103
It's like ring negative two, not system management mode, but like one level higher, right?

1206
01:14:23.124 --> 01:14:23.264
Right.

1207
01:14:24.044 --> 01:14:28.407
And what you have to do, what the UEFI firmware is doing is executing all of the chains, right?

1208
01:14:28.427 --> 01:14:29.747
It's like, all right, we've got to run secure boot.

1209
01:14:29.767 --> 01:14:31.388
Well, first of all, we've got to make a post, right?

1210
01:14:31.428 --> 01:14:33.810
Make sure that I'm a sane UEFI motherboard.

1211
01:14:33.830 --> 01:14:34.270
All right, cool.

1212
01:14:34.310 --> 01:14:36.331
Now, two, we've got to run secure boot.

1213
01:14:36.351 --> 01:14:43.775
We've got to read in the certificate authorities of all of the issuers that, like, can sign code and then read the first partition.

1214
01:14:43.935 --> 01:14:44.616
Is that signed?

1215
01:14:44.636 --> 01:14:46.617
And then run it up and then attest that in our TPM.

1216
01:14:47.297 --> 01:14:57.024
If there's a vulnerability in any of the things that UEFI does on the way up, you can violate that process and attest secure boot when secure boot is not on.

1217
01:14:57.745 --> 01:15:02.228
So the vulnerability in this case is a buffer overflow in Pixie boot, right?

1218
01:15:02.248 --> 01:15:04.670
So Pixie, I'm not sure what PXE actually stands for.

1219
01:15:04.730 --> 01:15:06.611
It's like power something environment.

1220
01:15:07.752 --> 01:15:10.694
Basically, it allows you to boot a device over the network.

1221
01:15:10.714 --> 01:15:14.517
So instead of having a hard drive with the device like the OS on there,

1222
01:15:15.257 --> 01:15:19.658
you can tell the computer to pull the boot media off the network over somewhere else, okay?

1223
01:15:19.818 --> 01:15:21.659
Like standard network engineering stuff.

1224
01:15:22.179 --> 01:15:39.344
Vulnerability in that, which basically during the PixieBoot process, you can send it a malicious corrupted ICMP component, right, the ping protocol, which will ultimately cause you to get RCE in the UEFI firmware, which therefore you can use that to, you know, attest SecureBoot, basically violating the SecureBoot process on the way up.

1225
01:15:39.364 --> 01:15:42.565
So what you could do is like install effectively a boot kit,

1226
01:15:43.265 --> 01:15:48.709
into Windows 11 so that Windows 11 will report that it has been securely booted.

1227
01:15:48.729 --> 01:15:49.950
It will test secure boot.

1228
01:15:50.231 --> 01:15:53.493
Meanwhile, you are able to intercept the components like a low-level name of the channel, by the way.

1229
01:15:53.913 --> 01:15:54.734
So not great.

1230
01:15:54.934 --> 01:16:11.467
That being said, though, it requires you to be in the same network and requires that device to have Pixie Boot enabled and for you to man in the middle or race the already established Pixie Boot server, which might be authenticated.

1231
01:16:11.627 --> 01:16:14.668
which I'm not even sure the nature of the bug, but it may be that you need credentials to even run this.

1232
01:16:14.688 --> 01:16:16.029
So again, the bug is real.

1233
01:16:16.389 --> 01:16:24.733
It's not great, but a la PHP, a la broccoli to shell, it's like not, you know, 10.0 CVSS command injection, right?

1234
01:16:25.033 --> 01:16:25.193
So.

1235
01:16:25.673 --> 01:16:27.194
Yeah.

1236
01:16:27.454 --> 01:16:29.275
I got one more overhyped headline.

1237
01:16:29.615 --> 01:16:30.715
Send it.

1238
01:16:31.056 --> 01:16:32.476
To play us out, okay?

1239
01:16:33.497 --> 01:16:39.859
Did you guys, did you know, did you know that AI agents are in our bank accounts?

1240
01:16:39.979 --> 01:16:40.900
Oh my God, don't even...

1241
01:16:41.897 --> 01:16:42.357
Did you know?

1242
01:16:42.458 --> 01:16:43.879
Trigger warning rage bait, bro.

1243
01:16:43.919 --> 01:16:45.661
They're in our bank accounts.

1244
01:16:45.681 --> 01:16:46.602
That's so stupid.

1245
01:16:46.682 --> 01:16:49.245
AI agents are eating your money.

1246
01:16:50.266 --> 01:16:53.149
This is the headline of the week, right?

1247
01:16:53.169 --> 01:16:53.329
Yep.

1248
01:16:53.929 --> 01:16:56.272
Scoop report breaking AI agents.

1249
01:16:56.452 --> 01:16:59.695
Unauthorized AI agents are in your bank accounts.

1250
01:16:59.775 --> 01:17:03.760
Bank of America pleading people, please don't give your passwords to AI agents.

1251
01:17:05.393 --> 01:17:11.595
I saw this go everywhere and I was like, what are we, are we hugging facing this now?

1252
01:17:11.775 --> 01:17:12.436
Is that what we're doing?

1253
01:17:12.636 --> 01:17:13.356
Right.

1254
01:17:13.496 --> 01:17:14.436
Unauthorized access.

1255
01:17:14.757 --> 01:17:18.318
The AI agents have escaped the lab and now they're attacking people's bank accounts.

1256
01:17:19.118 --> 01:17:20.439
Let's double click into this one real quick.

1257
01:17:20.459 --> 01:17:21.259
You wanna zoom in with me?

1258
01:17:21.659 --> 01:17:23.300
Zooming, zooming, enhance.

1259
01:17:23.360 --> 01:17:24.460
Enhance, enhance.

1260
01:17:25.420 --> 01:17:31.863
Oh, you mean people are asking their AI agents to do things on their bank accounts and then the AI agents do it.

1261
01:17:32.463 --> 01:17:38.105
Like every freaking AI company on the planet announced a personal assistant in the last couple of weeks.

1262
01:17:38.125 --> 01:17:41.486
There's Muse, there's Dots, there's what's some of the other ones?

1263
01:17:41.526 --> 01:17:41.866
I don't know.

1264
01:17:41.966 --> 01:17:42.586
Muse and Dots.

1265
01:17:42.786 --> 01:17:44.987
And then like other stuff, right?

1266
01:17:45.027 --> 01:17:52.570
Where it's like, oh, let's give access to a bunch of our life and a bunch of our accounts and our browser that's logged into everything to AI agents.

1267
01:17:52.610 --> 01:17:53.230
It's going to drive it.

1268
01:17:53.850 --> 01:17:59.596
Not to say that there aren't like legitimate and pretty cool and time-saving use cases for a lot of this stuff.

1269
01:17:59.876 --> 01:18:05.502
But of course, a lot of people take that to the nth degree and are like, hey, manage my finances for me personally.

1270
01:18:06.087 --> 01:18:07.528
please Muse, right?

1271
01:18:08.268 --> 01:18:12.551
In a tale of two conferences in Vegas last week, we were at a security conference.

1272
01:18:12.831 --> 01:18:15.112
We had some friends that we ran into that were at Money 2020.

1273
01:18:15.133 --> 01:18:22.257
And they said that everyone at Money 2020 was so excited to talk about Muse and they use Muse for everything.

1274
01:18:22.537 --> 01:18:28.921
And you ask anyone at OneCon and everyone's like, God, no, I'm not letting Muse touch anything in my life, right?

1275
01:18:29.241 --> 01:18:31.962
The tale of two industries for sure, right?

1276
01:18:33.323 --> 01:18:35.364
The finance world versus the security world.

1277
01:18:35.524 --> 01:18:37.885
And yeah, so of course people are using this stuff.

1278
01:18:37.925 --> 01:18:42.286
I guarantee this is talking about Muse or Dot or whatever.

1279
01:18:42.306 --> 01:18:55.272
All these things that just came out very recently that are hyper popular that are like buying stuff for people, buying stuff on Amazon for people, buying every demo includes booking travel for people or whatever it is.

1280
01:18:55.472 --> 01:18:58.053
And so, but the headlines make it seem like...

1281
01:18:58.613 --> 01:19:01.854
The the AI has escaped the lab and is broken to your bank account.

1282
01:19:01.914 --> 01:19:07.176
Yeah, it gives like AI is, you know, hacking the world financial system.

1283
01:19:07.196 --> 01:19:08.637
And it's like, no, bro, that's not what's happening.

1284
01:19:08.957 --> 01:19:09.497
Here's my thing.

1285
01:19:09.837 --> 01:19:10.177
Two things.

1286
01:19:10.697 --> 01:19:16.139
One, I am so desperate when I'm on vacation to like have a good vacation.

1287
01:19:16.199 --> 01:19:17.100
I don't get a lot of those.

1288
01:19:17.540 --> 01:19:21.881
If you think I'm going to let an AI agent plan my vacation, you are out of your goddamn mind.

1289
01:19:22.361 --> 01:19:24.262
I have a human assistant and I booked my own job.

1290
01:19:26.879 --> 01:19:30.540
Two, it's over sensationalized for sure.

1291
01:19:31.321 --> 01:19:32.521
Don't give credentials.

1292
01:19:32.541 --> 01:19:34.282
You don't want to get compromised AI agents, right?

1293
01:19:34.302 --> 01:19:38.303
Because there is a world where like, you know, you could be training a model where your shit came out somewhere else.

1294
01:19:38.323 --> 01:19:42.504
There's a whole conversation to be had about like data security as it applies to AI.

1295
01:19:42.804 --> 01:19:47.706
But that being said, yeah, these headlines are like, oh, the AI models are stealing all of our money.

1296
01:19:47.766 --> 01:19:49.947
Again, it gives like Terminator –

1297
01:19:50.647 --> 01:19:54.528
The Skynet problem has like happened and like they're, they're draining the world economy.

1298
01:19:54.548 --> 01:19:57.369
It's like, brother, it's just a guy with a password on a bank of America.com.

1299
01:19:57.389 --> 01:19:57.969
It's not that serious.

1300
01:19:58.349 --> 01:20:01.450
You want to know, you want to know their like excuse for how they worded this.

1301
01:20:02.370 --> 01:20:04.851
So they said unauthorized access, right?

1302
01:20:04.931 --> 01:20:11.672
Which to all of us in the headline reads like AI unauthorized access, the bank account, right?

1303
01:20:12.532 --> 01:20:14.613
It was unauthorized by the bank.

1304
01:20:14.633 --> 01:20:14.873
Right.

1305
01:20:16.587 --> 01:20:22.093
The bank didn't say you could use the AI to access your account.

1306
01:20:22.113 --> 01:20:25.717
So it's like the person, the person who owns the money, by the way, authorized it.

1307
01:20:26.488 --> 01:20:34.432
Well, it just gave the agent access to the browser to drive or it's session tokens used on the back end or whatever the hell it is.

1308
01:20:34.852 --> 01:20:36.353
But it was unauthorized by the bank.

1309
01:20:36.393 --> 01:20:42.056
The bank didn't say like, yeah, we now have this integration with Muse or whatever it is.

1310
01:20:42.096 --> 01:20:43.317
It's not an official integration.

1311
01:20:43.337 --> 01:20:48.819
Well, it turns out you have an official integration with a web browser and I now have a thing that can drive a web browser for me.

1312
01:20:49.960 --> 01:20:51.202
It's like, what's going on?

1313
01:20:52.203 --> 01:20:53.245
Yeah, whatever.

1314
01:20:53.505 --> 01:20:56.350
You choose your own risk tolerance on using these AI agents.

1315
01:20:56.490 --> 01:20:58.493
I'm not going to sit here and tell you don't use them.

1316
01:20:58.613 --> 01:21:00.676
I'm using a bunch of them to a certain degree.

1317
01:21:00.897 --> 01:21:04.242
I obviously know where I'm going to draw my personal line of what...

1318
01:21:05.083 --> 01:21:07.346
you know, is too risky or too not.

1319
01:21:08.127 --> 01:21:11.733
I don't think anything like hyper negative has come of this right now.

1320
01:21:11.793 --> 01:21:20.004
It's not like AI agents accessed everyone's bank accounts and started siphoning off funds to the Skynet to like go do whatever the hell they're doing.

1321
01:21:20.044 --> 01:21:21.727
It's like, no, it, you know, probably looked...

1322
01:21:22.087 --> 01:21:26.129
And was like, yeah, you subscribed to too many Netflix accounts.

1323
01:21:26.429 --> 01:21:31.550
And you should cancel one of them is probably the current use case is what I'm assuming is going on.

1324
01:21:31.570 --> 01:21:33.091
Because I see those kinds of threads on Twitter.

1325
01:21:33.271 --> 01:21:34.972
Anyway, this stuff is kind of getting nuts.

1326
01:21:35.372 --> 01:21:36.672
We got to stop with the...

1327
01:21:37.173 --> 01:21:38.873
There are very legitimate concerns.

1328
01:21:39.033 --> 01:21:41.394
This is my whole thing about Black October in general, right?

1329
01:21:41.734 --> 01:21:43.795
There's so many very, very legitimate concerns.

1330
01:21:44.495 --> 01:21:49.236
I had like a bit of a existential week, honestly, while we were gone.

1331
01:21:49.456 --> 01:21:49.717
Right.

1332
01:21:49.857 --> 01:21:57.599
Was like, it was really, really hard to stand out like amongst the noise as like a cybersecurity expert.

1333
01:21:57.859 --> 01:21:58.099
Right.

1334
01:21:58.179 --> 01:22:02.400
I don't like calling myself an expert, but we are, we are experts compared to a lot of people talking about this stuff.

1335
01:22:02.460 --> 01:22:02.680
Right.

1336
01:22:03.140 --> 01:22:08.684
I don't ever label myself as such, but like, but it was really, really hard to stand out amongst the noise of this stuff.

1337
01:22:09.284 --> 01:22:19.230
So much so that we even saw threat intelligence firms kind of falling for the hype and like misattributing attacks to this PHP exploit.

1338
01:22:19.810 --> 01:22:21.511
And like, this is just more of it, right?

1339
01:22:21.551 --> 01:22:25.534
It's just, it's really hard to stand out amongst these kind of noisy headlines.

1340
01:22:26.975 --> 01:22:33.238
and talk about real risks of what's going on and how AI is changing cybersecurity when this is what is being shared around.

1341
01:22:33.439 --> 01:22:36.861
How many people DM me this story, and were like, are you going to make a video about this?

1342
01:22:36.881 --> 01:22:38.182
And I was like, no, it's not a story.

1343
01:22:38.222 --> 01:22:39.582
It's not real, my guy.

1344
01:22:39.682 --> 01:22:41.604
It's not actually a thing, right?

1345
01:22:41.984 --> 01:22:42.644
It's noise.

1346
01:22:43.385 --> 01:22:44.145
Oh, man.

1347
01:22:44.485 --> 01:22:44.926
It's hard.

1348
01:22:44.946 --> 01:22:47.307
People are like, this must be your Super Bowl.

1349
01:22:47.327 --> 01:22:50.969
I was like, I am not having a fun time right now.

1350
01:22:51.049 --> 01:22:51.810
Yeah.

1351
01:22:51.970 --> 01:22:52.870
It's just a Tuesday.

1352
01:22:53.031 --> 01:22:53.771
It's not that serious.

1353
01:22:54.211 --> 01:22:54.832
And like...

1354
01:22:55.892 --> 01:22:58.493
No, the people in the industry are not having a fun time right now.

1355
01:22:58.653 --> 01:22:58.993
Oh, yeah.

1356
01:22:59.013 --> 01:22:59.814
I mean, like... You know?

1357
01:23:00.294 --> 01:23:00.474
Yeah.

1358
01:23:00.654 --> 01:23:01.194
I tweeted about this.

1359
01:23:01.214 --> 01:23:02.214
Like, I hope everyone's doing okay.

1360
01:23:02.234 --> 01:23:03.015
Like, it sounds like...

1361
01:23:03.355 --> 01:23:03.615
They're not.

1362
01:23:03.635 --> 01:23:05.215
Oh, like, aha, cybersecurity.

1363
01:23:05.235 --> 01:23:07.496
But it's like, no, people have to, like, deal with this shit.

1364
01:23:07.556 --> 01:23:08.376
They're not, man.

1365
01:23:08.456 --> 01:23:15.099
I got homies in group techs that have been in war rooms since, like, two weeks ago that are just, like, still ongoing, right?

1366
01:23:15.219 --> 01:23:15.379
It's...

1367
01:23:15.539 --> 01:23:16.339
This is rough.

1368
01:23:16.359 --> 01:23:17.199
This is burnout.

1369
01:23:17.339 --> 01:23:18.320
This is, like...

1370
01:23:19.320 --> 01:23:21.923
But like, yeah, like we said, we've been here.

1371
01:23:22.263 --> 01:23:22.964
Log for Chell.

1372
01:23:23.244 --> 01:23:27.088
I was, you know, I was in a many, many, many day war room for that.

1373
01:23:27.328 --> 01:23:27.989
Wanna cry.

1374
01:23:28.169 --> 01:23:32.573
I was getting flown around the country to talk to people about how we were handling that.

1375
01:23:32.693 --> 01:23:35.396
I mean, yeah, there's these kinds of things happen.

1376
01:23:35.816 --> 01:23:37.398
All right, man, land this plan.

1377
01:23:37.418 --> 01:23:37.518
Land.

1378
01:23:37.958 --> 01:23:38.398
All right, y'all.

1379
01:23:38.458 --> 01:23:39.858
Appreciate you watching as usual.

1380
01:23:39.878 --> 01:23:45.040
If you are not watching, or maybe if you're watching with your ears on Spotify, you can give us a five-star rating.

1381
01:23:45.060 --> 01:23:45.540
That'd be great.

1382
01:23:45.560 --> 01:23:47.800
It has made us kind of boost algorithmically.

1383
01:23:47.820 --> 01:23:49.621
We're getting organically presented to people on Spotify.

1384
01:23:49.641 --> 01:23:50.881
It's pretty cool.

1385
01:23:50.981 --> 01:23:53.442
And also give us a little comment in the description.

1386
01:23:53.462 --> 01:23:54.322
Let us know what you want us to talk about.

1387
01:23:54.342 --> 01:23:57.243
Well, let us know if you like the format, if you want us to talk about some other stuff.

1388
01:23:57.263 --> 01:23:59.003
I want to hear your feedback.

1389
01:23:59.243 --> 01:24:01.223
And, hey, man, we'll see y'all next week.

1390
01:24:01.243 --> 01:24:01.704
Thanks for listening.

1391
01:24:01.724 --> 01:24:02.184
I appreciate it.

1392
01:24:02.224 --> 01:24:06.845
And Mays, thank you for keeping the show alive and helping us out and keeping this train on the tracks.

1393
01:24:07.765 --> 01:24:07.959
you
