Can SOC Operators Explain their Decisions while Triaging Alarms? A Real-World Study

Collection of recent arXiv papers (Apr 27, 2026) on security, privacy, and trustworthy ML/AI systems. Key contributions include: an empirical SOC study showing analysts make correct triage decisions in 83% of cases but provide correct justifications only 39%; a blockchain-anchored, tamper-evident fraud-detection system with cryptographically verifiable decision trails (F1=0.895, sub-25 ms inference, <$0.01/tx on L2); Sovereign Agentic Loops (SAL) architecture that emits structured intents and an Evidence Chain to prevent unsafe LLM-driven executions (blocks 93% of unsafe intents, adds ~12.4 ms