LATIDIA · Ciberseguridad

VOTO: Detección de marcas de agua verificables y ajenas para modelos de lenguaje grandes

arXiv: 2604.27666v2Tipo de anuncio: reemplazar Resumen: La regulación requiere que el texto generado esté marcado y sea detectable, pero las marcas de agua de clave simétrica dejan la detección al proveedor de mantenimiento de clave. Estudiamos provenanc de primera mano

La noticia

arXiv:2604.27666v2 Announce Type: replace Abstract: Regulation requires generated text to be marked and detectable, but symmetric-key watermarks leave detection to the key-holding provider. We study first-party provenance audit (FPA), where an organization examines sensitive content in its custody. Such an audit should not disclose the audited text and should provide a verifiable result and checkable record. We present VOW, which uses a verifiable oblivious pseudorandom function to evaluate the detector's keyed component without revealing its inputs to the detection service, while the provider retains the watermarking key. A signed setup record published before use fixes the public key and detector specification, and a proof binds returned evaluations to that key.