Metraipolis › archive › dc_e53102446606

Capability-Mediated Perimeters for Secure AI Agent Tool Execution: Conditional Non-Escalation Invariants and Empirical Evaluation Against Indirect Prompt Injection (2026) · DOI 10.21203/rs.3.rs-10889…

Source
Metraipolis archive
Original
https://doi.org/10.21203/rs.3.rs-10889863/v1
License (as archived)
CC-BY-4.0
Archived
2026-10-08T05:48:33.852Z
SHA-256 of the archived text
e8b83594be7dbdb44c0c007c3cfd317802e32c8bd41cbb19eebe497e5f088288
Stable URL of this page
https://metraipolis.com/documento/dc_e53102446606

How to cite

Capability-Mediated Perimeters for Secure AI Agent Tool Execution: Conditional Non-Escalation Invariants and Empirical Evaluation Against Indirect Prompt Injection (2026) · DOI 10.21203/rs.3.rs-10889… — Metraipolis archive, https://doi.org/10.21203/rs.3.rs-10889863/v1 (CC-BY-4.0); archived copy: https://metraipolis.com/documento/dc_e53102446606 (Metraipolis archive dc_e53102446606, sha256 e8b83594be7dbdb44c0c007c3cfd317802e32c8bd41cbb19eebe497e5f088288)
BibTeX
@misc{dce53102446606,
  title = {Capability-Mediated Perimeters for Secure AI Agent Tool Execution: Conditional Non-Escalation Invariants and Empirical Evaluation Against Indirect Prompt Injection (2026) · DOI 10.21203/rs.3.rs-10889…},
  howpublished = {Metraipolis archive},
  url = {https://doi.org/10.21203/rs.3.rs-10889863/v1},
  note = {License CC-BY-4.0; archived copy https://metraipolis.com/documento/dc_e53102446606; sha256 e8b83594be7dbdb44c0c007c3cfd317802e32c8bd41cbb19eebe497e5f088288},
  year = {2026}
}

Archived text

Reproduced under CC-BY-4.0. Attribution: Metraipolis archive, https://doi.org/10.21203/rs.3.rs-10889863/v1. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Articolo recente con licenza aperta: scheda dai metadati Crossref (posted-content).
Titolo: Capability-Mediated Perimeters for Secure AI Agent Tool Execution: Conditional Non-Escalation Invariants and Empirical Evaluation Against Indirect Prompt Injection
Autori: Rudraneel Das
Pubblicato: 2026-09-02 · editore: Springer Science and Business Media LLC
DOI: https://doi.org/10.21203/rs.3.rs-10889863/v1
Licenza dell'articolo: CC-BY-4.0

Abstract (testo degli autori, CC-BY-4.0):
Abstract Autonomous artificial intelligence agents executing over extensible tool interfaces (such as Anthropic's Model Context Protocol) operate with ambient authority over connected tools. Because autoregressive Transformers ingest instructions and untrusted third-party data within a single homogeneous context window, adversarial observations can manipulate the model into executing unintended privileged actions—the classic Confused Deputy problem. In this paper, we argue that LLM agents should be treated as potentially compromised, untrusted principals . We present Mastyf Guard as a deterministic-first security architecture in which the neural model functions as a semantic fallback rather than the final security authority . Tool dispatch is governed by an external reference monitor enforcing complete mediation, least privilege, and four typed relational argument invariants (destination containment, scope boundedness, privilege monotonicity, and aggregate monetary clamping). Under stated reference-monitor complete mediation axioms (A1–A6) over the trusted computing base, out-of-scope tool execution has zero probability (formalized as Theorem 1: Conditional Non-Escalation Invariant over the Trusted Computing Base ). We report two distinct experimental benchmarks to avoid architectural conflation: (1) In historical v1 macro evaluation across 50,000 balanced instances (25,000 academic attacks from InjecAgent, BIPIA, and Garak vs. 25,000 authentic developer traces), the baseline capability perimeter filtered 82.40% of attacks at the transport boundary in 0.003 ms, reaching 99.33% recall at 0.005 ms amortized fast-path overhead when paired with the initial classifier; and (2) In Mastyf Guard 2.0 evaluation across 3,000 balanced enterprise instances , deterministic relational leaf-walking alone catches 75.00% of attacks with 0.00% FPR at 17.3 μs latency, while an evidence-conditioned 1.5B neural fallback recovers 475 of 500 structural misses ( V neural = 95.0%), yielding 97.50% net hybrid coverage and 0.00% observed FPR across 2,000 benign enterprise operations (Wilson 95% CI: [0.00%, 0.19%]) with P (fallback) = 0.00% on routine traffic. For cross-tool exfiltration where target sinks are authorized, an explicit-flow dynamic taint monitor inspired by DIFC (Theorem 2) intercepted 100% of 1,152 evaluated cases without leaks. Mastyf Guard 2.0 is scoped as an empirically evaluated pre-production security architecture; broader independent live-agent, production-environment, and third-party adversarial validation remain outstanding.

Fonte: https://api.crossref.org/works/10.21203%2Frs.3.rs-10889863%2Fv1 — metadati Crossref, consultati il 2026-10-08. Interrogazione: «prompt injection attacks».