LATIDIA · Ciberseguridad

Una evaluación de las capacidades de comprensión semántica de modelos de lenguaje grandes para cargas útiles de ataques web

arXiv: 2610.06507v1Tipo de anuncio: nuevo Resumen: Los servicios de visión por computadora entregados a través de interfaces web y API procesan solicitudes textuales para la adquisición de recursos de imágenes, la configuración de tareas de inferencia y la gestión de resultados

La noticia

arXiv:2610.06507v1 Announce Type: new Abstract: Computer vision services delivered through Web interfaces and APIs process textual requests for image-resource acquisition, inference-task configuration, and result management, making Web attack-payload analysis relevant to their deployment security. Large language models (LLMs) can identify payload types and explain attack intent. However, existing studies generally treat payload analysis as a single-layer classification task and lack both a systematic assessment of how deeply LLMs understand payloads and an evaluation benchmark dedicated to the depth of semantic understanding of Web attack payloads. We construct PayloadSemBench, a four-layer semantic evaluation benchmark that operationalizes payload understanding across measurable tasks and comprises 240 payloads. Its ground truth was established through