LowResearchPeer-reviewed
Generative Red Teaming for Robust Network Intrusion Detection: A Constrained Adversarial Approach
- Published
- Record updated
Summary
Researchers introduce PerturbLite, a generative red teaming framework for assessing ML/DL-based network intrusion detection systems (NIDS) under realistic constraints. Its autoencoder-based model produces minimally perturbed, functionality-preserving adversarial packets that evade detection, often altering less than 1% of the original content. The authors report superior evasion performance over existing adversarial methods on real-world intrusion datasets and say the framework exposes systemic NIDS vulnerabilities.
Related items
- InfoOut of thin air: An adversarial attack method against human pose estimation networkSimilar attack · Elsevier Security Journals
- InfoAdversarial Attacks on Vision-Language Multimodal Systems: A SurveySimilar attack · OpenAlex (peer-reviewed AI security)
- MediumGHSA-25rm-9wvm-m38v: falcon-rust: Discrete-Gaussian sampler runs at 2^-33 precision, far below Falcon's requirement (reference KAT vectors no longer pass)Similar attack · GitHub Advisory Database
- InfoAdversarial Malware Detection Using Feature Contribution Difference Analysis with Kernel SHAPSimilar attack · OpenAlex (peer-reviewed AI security)
- InfoDensity Feedback Control for Transferability of Generative Sparse Adversarial AttacksSimilar attack · Arxiv (cs.RO + cs.CV security)