analiza aiqbench
This research addresses a critical security challenge where autonomous agents coordinate distributed reconnaissance across thousands of identities, effectively hiding individual malicious intent within benign-looking traffic. The study provides a rigorous black-box benchmark comparing various statistical and graph-based detectors to identify these collective behaviors without access to internal system states. While the core methodology involves machine learning classifiers, the primary subject is cybersecurity defense against emerging agentic threats, making it relevant for AI safety and policy rather than general ML advancement.
SwarmReconGuard: Black-Box Detection of Distributed Collective Reconnaissance by Individually Benign-Looking Agent Populations
arXiv:2610.09138v1 Announce Type: cross Abstract: Autonomous and agentic clients can distribute reconnaissance across many identities so that each request remains valid, low-rate, and benign-looking while the population collectively acquires broad system knowledge. We formalize this threat as Distributed Collective Reconnaissance (DCR) and present SwarmReconGuard, a reproducible black-box benchmark in which the defender observes only service-boundary telemetry. The Docker-isolated study evaluates 11 benign and attack behaviors across 10-10,000 virtual identities, comprising 440 test runs and 3,666,300 requests, with complete telemetry integrity. We compare semantic, Gaussian, conditional, graph, kernel, hybrid, and CUSUM-based detectors. Gaussian likelihood-ratio detection achieves 100$\%$ detection with 0$\%$ observed false positives on known attacks but only 3$\%$ on unseen policies. CUSUM yields 36.1$\%$ overall detection at 1.25$\%$ false positives, while hybrid CUSUM reaches 85.7$\%$ detection with 0$\%$ observed false positives at 10,000 identities. Results expose a major policy-generalization gap and motivate exposure-aware, scale-aware defenses.
Abstract:Autonomous and agentic clients can distribute reconnaissance across many identities so that each request remains valid, low-rate, and benign-looking while the population collectively acquires broad system knowledge. We formalize this threat as Distributed Collective Reconnaissance (DCR) and present SwarmReconGuard, a reproducible black-box benchmark in which the defender observes only service-boundary telemetry. The Docker-isolated study evaluates 11 benign and attack behaviors across 10-10,000 virtual identities, comprising 440 test runs and 3,666,300 requests, with complete telemetry integrity. We compare semantic, Gaussian, conditional, graph, kernel, hybrid, and CUSUM-based detectors. Gaussian likelihood-ratio detection achieves 100$\%$ detection with 0$\%$ observed false positives on known attacks but only 3$\%$ on unseen policies. CUSUM yields 36.1$\%$ overall detection at 1.25$\%$ false positives, while hybrid CUSUM reaches 85.7$\%$ detection with 0$\%$ observed false positives at 10,000 identities. Results expose a major policy-generalization gap and motivate exposure-aware, scale-aware defenses.
Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
arXiv:2610.09138 [cs.CR]
(or arXiv:2610.09138v1 [cs.CR] for this version)
https://doi.org/10.48550/arXiv.2610.09138 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 241 cuvinte
Subiecte conexe
Improving CoT Monitorability of Evaluation Awareness via Verbalization Training
LessWrong · 2 oct. 2026
Brief independent investigation of agents’ behavior, reasoning and collaboration in the OpenAI / Hugging Face hacking incident
METR · 26 aug. 2026
Localize-and-Detect: Auditing Task-Level Poisoning in Instruction-Tuned Models
arXiv · 6 oct. 2026
Closing the Binary Blind Spot: Domain-Knowledge Injection for LLM Malware Triage in npm Packages
Databricks · 16 oct. 2026