Firmware rehosting enables embedded firmware to be executed outside its original hardware environment for dynamic analysis/fuzzing. However, rehosting microcontroller firmware often requires substantial manual effort to identify hardware dependencies, model peripheral behavior, and diagnose execution failures. Recent advances in large language models (LLMs) offer an opportunity to automate parts of this process by reasoning over firmware binaries, static-analysis results, execution traces, documentation, and existing rehosting tools.

Objectives

• Automate rehosting tasks: Investigate how LLMs can assist with identifying hardware dependencies, interpreting execution failures, and generating peripheral models, hooks, or rehosting configurations.
• Develop an AI-assisted workflow: Integrate LLM-based assistance with existing binary-analysis and firmware-rehosting techniques, with a focus on ARM Cortex-M firmware.
• Evaluate generalization: Apply the approach to multiple firmware targets with different hardware and peripheral characteristics.
• Measure analyst effort: Compare the AI-assisted approach against an existing manually developed rehosting baseline and evaluate whether it reduces the time and expert effort required to obtain a working rehosting environment.

Requirements

• Embedded systems and hardware security.
• Experience in reverse engineering and low-level programming.
• Familiarity with ARM architectures is advantageous.
• Mindset to learn the additional skills.