TAS-GNN: A Status-Aware Signed Graph Neural Network for Anomaly Detection in Bitcoin Trust Systems

2026-03-17T07:23:56Z70d3b75cf04f5928f04f917a9e0aa2160efdd93649c160c29cd4004c7e40ed7d
adversarial-mlagent-isolationanomaly-detectionbackdoor-removalbenchmarkbitcoincti-REALMdifferential-privacygraph-neural-networkimage-encryptioninference-optimizationjson-sanitizationkv-cachemodel-backdoorsprivacy-enhancing-technologiesprivilege-separationprompt-injectionquantum-rngrandom-number-generatorred-teamingsigned-graphssuffix-jailbreaksybil-attacksynthetic-dataweb-of-trust

What happened

Collection of recent security- and privacy-relevant CS papers (arXiv 2026-03-17). Highlights: TAS-GNN — a topology- and status-aware signed GNN for anomaly detection in Bitcoin Web-of-Trust networks (improves detection of Sybil/camouflaged fraud); Prefix-Shared KV Cache (PSKV) — an inference optimization that reuses a prefix KV cache to accelerate suffix jailbreak candidate evaluation (reported ~40% runtime and ~50% peak memory reductions; code to be released), which can make suffix jailbreak red-teaming (and offensive testing) much cheaper; OpenClaw agent privilege separation — a structural,

Why it matters

A reviewed impact interpretation has not been published for this record.

Evidence and limitations

Source ID
arxiv_cs_cr
Record identifier
70d3b75cf04f5928f04f917a9e0aa2160efdd93649c160c29cd4004c7e40ed7d
Enrichment time
2026-03-17T07:23:56Z
AI-assisted enrichment
Yes

This record may overlap with other records. Its enrichment can be incomplete or wrong, and machine assistance was used. Validate consequential decisions against the linked source and your own environment.