Asymmetric Phase Coding Video Watermarking

A new video watermarking approach, Asymmetric Phase Coding Video Watermarking, replaces the usual shared-secret model with a public-key verification flow. In practical terms, the signer embeds a full Ed25519 signature into the phase spectrum of the chroma plane, while anyone with the 32-byte public key and per-video metadata can verify it offline. That makes the system more like a cryptographic signature than a conventional watermark pipeline.

The big technical point for developers is that the method is training-free and does not rely on a learned extractor or generator. The design combines a run-length temporal layout, payload-free geometric recovery for scale/rotation/translation, and a closed-loop signing step that tunes embedding strength through self-verification. The payload is 1024 bits of signed message plus error correction, which is notably larger than many learned watermark payloads.

The authors tested the system on 1,000 uncurated real-world clips and report successful verification on 99.3% of them, with zero false accepts against the wrong public key in 1,000 attempts. An attack-aware acceptance gate is said to preserve marks through H.264 re-encoding at 100% and 50% rescaling on gated clips at 97.4%. The signature also survives a real display-and-capture loop, which is a useful benchmark for production-facing media workflows.

For game teams, the relevance is strongest anywhere video authenticity matters: trailers, UGC pipelines, anti-tamper checks, and provenance tracking for captured footage. The asymmetric model is especially interesting because it...