LATIDIA · Ciberseguridad

Confíe, pero valide el instrumento: audite los planes de verificación RTL generados por IA en los proxies de regresión de seguridad creados

arXiv: 2609.19844v1Tipo de anuncio: nuevo Resumen: los planes de verificación RTL generados por IA pueden satisfacer un esquema de proveedor pero fallar en el límite de la ejecución confiable. Presentamos SecTB-RTL, un marco auditable que cubre 31 ta

La noticia

arXiv:2609.19844v1 Announce Type: new Abstract: AI-generated RTL verification plans can satisfy a provider schema yet fail at the boundary to trusted execution. We present SecTB-RTL, an auditable framework covering 31 tasks and 124 authored hardware-security regressions. A deterministic non-AI baseline killed 36, 75, and 78 mutants at increasing resource limits. The first confirmatory run (C1-R2) failed before model execution because the provider rejected its response schema. After a schema-only repair made without viewing outcomes, a separately frozen follow-up run (C1-R3) completed 1,860 calls. The provider accepted 1,857 responses, but only nine passed the production semantic validator. The generation and execution rules did not match. We therefore preserve the run as