Reproduce “Robust Intrusion Detection System with Explainable Artificial Intelligence” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Betül Güvenç Paltun, Ramin Fuladi, Rim El Malki
Ericsson Research developed an Intrusion Detection System that integrates Explainable Artificial Intelligence (XAI) using SHAP values to detect adversarial attacks against machine learning models in Open Radio Access Network (O-RAN) environments. This approach identifies anomalous feature importance patterns caused by adversarial manipulation, enabling real-time detection with over 90% accuracy and automatically restoring IDS performance.
Key insights
- Adversarial attacks on ML models, even with subtle perturbations, can significantly alter the feature importance rankings, making these changes detectable through XAI methods.
- Monitoring the distribution of SHAP values allows for real-time identification of adversarial manipulations, as they deviate from established normal behavior profiles.
- SHAP consistently outperforms other Explainable AI methods like LIME and permutation importance in accurately detecting adversarial attacks on intrusion detection systems.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.