hn.today

LLM Assisted Reverse Engineering

slowchronicle.com302 points77 comments
Screenshot of LLM Assisted Reverse Engineering

A longtime hardware and software reverse engineer describes using LLM agents to make binary analysis far faster. Their toolkit centers on DeepSeek, Ghidra with an MCP bridge, and command-line tools such as radare2 and kuna. The agent first identifies a binary’s platform and structure, then examines functions, control flow, and imports. It records discoveries in Markdown files so both the human and agent can maintain context across long investigations. When static analysis stalls, the workflow shifts to debugging, breakpoints, logging probes, desktop interaction, or emulators. The human also reviews the agent’s reasoning and redirects it when it fixates on an unproductive path.

The most important discipline is hypothesis-driven debugging: record a proposed explanation, design an experiment that can decisively test it, log the outcome, and repeat rather than guess indefinitely. Dedicated skills for reverse-engineering tools and for eliciting crucial user input help agents work more effectively. Python with UV supports quick probes and prototypes; Rust is preferred for finished modifications because its compiler gives agents rapid, strict feedback. The author says this approach has enabled projects that would once have taken years, including mapping device protocols, modifying synthesizer and ebook-reader firmware, repairing a printer driver, and building custom applications. LLMs do not replace expertise: experience is still needed to steer investigations, but inexpensive, capable models turn a slow hobby into a much faster process of exploration and learning.

Read on slowchronicle.com77 comments on Hacker News

Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.

More in AI

Mapping every neuron in Qwen3-8B

Mapping every neuron in Qwen3-8B

Researchers found individual neurons in Qwen3-8B that reliably steer its replies into distinct personas or styles, including pirate, German-speaking, and childlike voices. Nudging these neurons changed outputs across many questions without prompting or fine-tuning, revealing how specific behaviors can be localized and edited inside a language model. (fermionresearch.com)

ChatGPT for SNES

ChatGPT for SNES

ChatGPT for SNES runs a 2.2-million-parameter language model on the Super Nintendo’s Super FX 2 chip, with an optional Wikipedia fact book. The unofficial experiment can run in an emulator or on a flash-cart-equipped console, but its tiny model often gives incorrect answers. (bxfoundry.com)

The daily digest

Today's best Hacker News stories, summarized and screenshotted, one email a day.