The classifier judged this item not relevant to AI security, so it is left out of lists, feeds, the API and dataset releases. If that is wrong, report a correction on this page.
InfoResearchPeer-reviewed
ProvPlug: LLM-Driven Plugins for Bridging Provenance Graph Host Intrusion Detection Systems and Graph Neural Networks
- Published
- Record updated
Summary
ProvPlug is an LLM-driven plugin framework that improves compatibility between data provenance graphs (DPGs) and graph neural networks (GNNs) for host intrusion detection. It combines two plugins, one adapting DPGs to GNNs and one adapting GNNs to DPGs, replacing previously manual specifications. Evaluated on eight state-of-the-art provenance graph host intrusion detection systems (PIDSs) across eight public datasets, it significantly improved their detection performance and outperformed all specification-based baselines.