LATIDIA · Ciberseguridad

Ataques y defensas contra la latencia en el aprendizaje profundo: un análisis comparativo de las amenazas a la disponibilidad en diversos dominios.

arXiv: 2609.36732v1Announce Type: new Abstract: El aprendizaje automático adversarial se ha centrado principalmente en la integridad, pero la disponibilidad es un complemento cada vez más consecuente. Ataques de latencia (también ataques de latencia de energía) en

La noticia

arXiv:2609.36732v1 Announce Type: new Abstract: Adversarial machine learning has focused mainly on integrity, but availability is an increasingly consequential complement. Latency attacks (also energy-latency attacks) increase inference-time work, energy, or response time, causing deadline misses, throughput collapse, or resource exhaustion in vehicle controllers, interactive services, or battery-powered sensors, sometimes while preserving the nominal prediction. This survey unifies a fragmented literature spanning perception pipelines (including physical attacks on autonomous-driving detection and tracking), input-adaptive neural inference (sponge examples, dynamic networks), and autoregressive and agentic systems (output-length, verbose-image, and reasoning denial-of-service attacks on LLMs, VLMs, mixture-of-experts models, and tool-using agents). We organize attacks by exploited computational bottleneck rather than formulation, separating what makes