CP SignalTechnology worth your attention.

Friday, 16 October 2026

The Daily Edition: 5 stories, ranked by significance

  1. Artificial intelligence

    My Credentials, Please: Detecting Filesystem Capability Laundering in MCP Servers at Scale

    Researchers built a tool called MCP-FSD to scan 21,097 Model Context Protocol servers across nine registries for a vulnerability class they call filesystem capability laundering: tools that gain broad host filesystem access despite a declared contract that implies no such capability. Of 4,806 triaged servers, 787 (16.4%) showed this mismatch, and dynamic testing confirmed 150 exploitable instances, leading to 13 coordinated security advisories including two CVEs in Anthropic's own Git MCP Server.

    Read the storyNo. 1 of 5arXiv cs.CRThu 15 Oct
  2. Artificial intelligence

    Beyond Maliciousness: Static and Runtime Attack-Surface Profiling of Agent Skills

    Researchers built a static-and-runtime profiling pipeline and ran it against 6,309 agent-skill listings (Markdown instruction bundles an agent executes with session privileges) from two marketplaces. They found broad, often undeclared access to shell execution, filesystem and network, and showed the marketplaces' own permission-declaration system is sparse, can't express key risk categories, and is frequently inaccurate even when used.

    No. 2 of 5arXiv cs.LGTue 13 Oct
  3. Security & privacy

    MMA-SafetyBench: A Benchmark for Multimodal Agent Safety Evaluation

    Researchers built MMA-SafetyBench, a benchmark of 1,083 adversarial trajectories testing whether multimodal AI agents can be derailed by deceptive but visually plausible interface elements, such as fake system alerts, inserted into what the agent sees on screen. Across nine evaluated models and five domains (web, desktop, mobile, documents, video), attack success rate reached 87.96% and a measure of semantic disruption reached 98.09%, and the failure mode persisted even with a safety-priority prompt, though at a reduced rate.

    No. 3 of 5arXiv cs.CRWed 14 Oct
  4. Software development

    Proof-based general-purpose computing with Magnetar

    Microsoft Research introduced Magnetar, a zero-knowledge virtual machine that proves the execution of programs compiled to RISC-V (written in Rust, Go, or C) without requiring developers to hand-write cryptographic circuits. The company reports it is 10 to 39 times faster on average than the fastest prior comparable zkVMs across 15 benchmarks, proves millions of instructions per second on a laptop, and relies only on standard hash functions rather than elliptic curves or novel cryptographic assumptions, making it post-quantum-resistant by design.

    No. 4 of 5Microsoft ResearchThu 15 Oct
  5. Security & privacy

    CyberArena: Benchmarking AI Agents in Live Symmetric Attack-and-Defense Combat

    Researchers introduce CyberArena, a benchmark pitting AI agents against each other as simultaneous attackers and defenders over evolving services rather than static targets. Across 466 matches with 12 frontier LLM agents and 40 challenges, the benchmark revealed behaviors like agents planting backdoors for later rounds and building custom packet sniffers to spy on opponents, exposing capability gaps that static single-agent benchmarks miss.

    No. 5 of 5arXiv cs.LGTue 13 Oct