Pseudorandom Codes Enhance AI Watermarking with Provable Cryptographic Security
This research introduces novel pseudorandom error-correcting codes, a cryptographic primitive for provably secure and tamper-resistant watermarking of generative AI outputs.
