Thank you for your interest in contributing to AeroMind! This project is a research artifact accompanying the RAID 2026 paper. We welcome contributions that improve reproducibility, extend the evaluation, or add defensive countermeasures.
- Bug reports — open a GitHub Issue with a minimal reproduction case
- New attack scenarios — extend the
attacks/framework with new threat vectors - New defense mechanisms — add to
uavsys/memory/defense.py - Additional LLM backends — extend
uavsys/llm/ollama_client.py - Documentation improvements — fix typos, clarify setup steps, add examples
git clone https://github.com/OdatSec/AeroMind.git
cd AeroMind
python -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt- Fork the repository and create a feature branch:
git checkout -b feat/your-feature - Keep attack implementations isolated in
attacks/— never call attack code fromuavsys/ - Add a named config to
configs/baseline_configs.yamlfor any new experiment - Run the baseline (
B0) before and after changes to confirm no regressions - Open a Pull Request with a clear description of what was changed and why
If your contribution includes a new attack vector or vulnerability, please follow responsible disclosure norms and read SECURITY.md before opening a PR.
AeroMind · RAID 2026