2026-08-08, 22:15–22:45, Main tent
How useful are current generation LLMs for reverse engineering? Turns out, very!
A few weeks ago I stumbled over an old reverse engineering project I shelved due to lack of time and grit and thought "hey it can't hurt to throw my notes into claude and see what comes out". One full vendor license bypass and several broken crypto implementations later the project had grown into a full teardown and that's what I'm currently working on. (I won't mention the hardware on the record but I've talked about it in the past)
This is going to be more of a campfire talk - I'm going to talk for maybe five minutes and then I'm going to be fielding questions for however long it takes. I'm especially interested in how others have been using LLMs and other AI techniques for reverse engineering, because I've been living under a rock when it comes to infosec and damn are these things useful for RE work (What's a Hough transform? Well I learned this week). Also I'd like to hear about how traditional approaches are doing because it turns out five minutes of staring at the right bytes can still beat one day of clauding around.
There's going to be cool graphs and stories that embarrass both me and the vendor.