I have read through the curl vulnerability reports... poor guy. Everything is AI slop, including the discussion. Fuck these people.
| Github | https://github.com/nullt3r |
| Substack | https://safemode.substack.com |
Nullt3r
- 321 Followers
- 199 Following
- 141 Posts
hacker, tinkerer
this is similar to cpu attacks but so high level: side-channel attacks against LLMs https://www.schneier.com/blog/archives/2026/02/side-channel-attacks-against-llms.html
Side-Channel Attacks Against LLMs - Schneier on Security
Here are three papers describing different side-channel attacks against LLMs. “Remote Timing Attacks on Efficient Language Model Inference“: Abstract: Scaling up language models has significantly increased their capabilities. But larger models are slower models, and so there is now an extensive body of work (e.g., speculative sampling or parallel decoding) that improves the (average case) efficiency of language model generation. But these techniques introduce data-dependent timing characteristics. We show it is possible to exploit these timing differences to mount a timing attack. By monitoring the (encrypted) network traffic between a victim user and a remote language model, we can learn information about the content of messages by noting when responses are faster or slower. With complete black-box access, on open source systems we show how it is possible to learn the topic of a user’s conversation (e.g., medical advice vs. coding assistance) with 90%+ precision, and on production systems like OpenAI’s ChatGPT and Anthropic’s Claude we can distinguish between specific messages or infer the user’s language. We further show that an active adversary can leverage a boosting attack to recover PII placed in messages (e.g., phone numbers or credit card numbers) for open source systems. We conclude with potential defenses and directions for future work...
after nearly 4 years, I have finally released massmap (JFScan successor): https://github.com/nullt3r/massmap
Deviant OllamThank you to everybody who sent me the news stories and posts about the hacked crosswalk buttons in Silicon Valley and Seattle. Maybe we'll see them elsewhere, too, soon? I'm betting there may be other cities whose DOT chose weak passwords for systems like this.
If you haven't heard what's up, I did a super short video about this...
The Crosswalk Button Hack News Story
Practical HTTPS Interception https://blog.thc.org/practical-https-interception
Practical HTTPS Interception
TL;DR: An attacker can trick Let's Encrypt (LE) to issue new TLS certificates for any domain that the attacker intercepts traffic for. The attacker can then decrypt the TLS traffic. This one thing that TLS is supposed to prevent from happening. The f...
Is anyone actually using Burp AI? It seems like a feature added just in case the models become more useful in the future.
Updated the list of open-source reporting tools:
- https://github.com/blacklanternsecurity/writehat
- https://github.com/pwndoc/pwndoc
- https://github.com/Anof-cyber/APTRS
- https://github.com/Syslifters/sysreptor
- https://www.defectdojo.org/
Do you know more?
