LATIDIA · Ciberseguridad

Bloqueo combinatorio para motores de IA resistentes a troyanos mediante filtrado espectral asistido por hash

arXiv:2610.33437v2 Anuncio Tipo: reemplazar Resumen: Ataques de puerta trasera de activación múltiple, en los que coexisten mecanismos de activación heterogéneos (por ejemplo, parche, mezcla, distorsión y la marca de agua espía de dominio latente) dentro de un único punto de acceso.

La noticia

arXiv:2610.33437v2 Announce Type: replace Abstract: Multi-trigger backdoor attacks, in which heterogeneous trigger mechanisms (e.g., patch, blend, warping, and the latent-domain Spy-Watermark) coexist within a single poisoned model, cannot be addressed by existing shortcut-based defenses that assume a single, static attack mechanism. Current multi-trigger remedies (BU+FT, BU+ANP, and BU+ABL) rely on backdoor unlearning, which incurs high training costs and loses effectiveness at low poisoning ratios. This paper proposes Combinatorial Lock, a training-free, reference-calibrated defense that reduces multi-trigger identification to a combinatorial problem over binary activation signatures. This formulation is the key novelty: it yields a multi-trigger defense that is training-free, trigger-oblivious, and directly implementable in hardware, since its decision rules can