XAI and Statistical Analysis for Reliable Intrusion Detection in the UAVIDS-2025 Dataset: From Tree to Hybrid and Tabular DNN Ensembles

Collection of recent security research papers covering multiple emerging risks and defenses: explainable ML for UAV intrusion detection (UAVIDS-2025) identifying feature-level causes of Wormhole/Blackhole misclassifications; AgentTrap, a dynamic benchmark showing third‑party LLM skills can embed malicious runtime workflows that evade simple jailbreak detection; a Volatility 3 memory‑forensics framework for Go malware that reliably recovers runtime artifacts (C2 endpoints, keys, persistence) from samples like BRICKSTORM/Obscura/Pantegana; ExploitBench, a capability‑graded benchmark on 41 V8bugs