
Using AI to map an unfamiliar attack surface
Build a useful map of a system while keeping model suggestions separate from observed behavior.
Exploring vulnerabilities with AI, and going deep through reverse engineering in my independent work. Field notes on how software really behaves.

AI helps uncover a thread. Reverse engineering follows it into what the software actually does.

Methods for tracing a lead from the first observation to evidence you can examine.

Build a useful map of a system while keeping model suggestions separate from observed behavior.

Follow the clues from observed behavior into the implementation beneath it.

Separate an interesting signal from a vulnerability you can reproduce and explain.
Follow the thread through a binary.
An AI-assisted Ghidra workflow for discovering potential weaknesses in binaries. It connects decompiled code, assembly, and call context to evidence-linked findings, while keeping verification explicit.

I publish as Seichs and mainly work on AI-powered vulnerability research. In my independent work, I specialize in reverse engineering: looking beneath the interface to understand what software actually does.

Interested in my research or in working together? You can find my public work and current profile under Seichs.
View Seichs on GitHub ↗