H.264/AVC is a widespread standard for high definition video (HD) for example DVD and HD videos on the internet. To prevent unauthorized modifications, video authentication can be used. In this paper, we present a cryptanalysis of a H.264/AVC video authentication scheme proposed by Saadi et al. [1] at EUSIPCO 2009. Our result will prevent situations where newer schemes are developed from the scheme thus amplifying the flaw. The designers claimed that the scheme can detect modifications on watermarked video. However, we show that an attacker can modify the watermarked video and compute a valid watermark such that the recipient will retrieve a watermark from the modified watermarked video that will match what the recipient computes during video authentication check. Thus, the recipient will think the tampered video is authentic. The first main problem of the scheme is its use of hash functions for watermark generation. Since hash functions are public functions not depending on any secret, the attacker can modify the watermarked video and feed this through the hash function to compute a new watermark. The second problem is that it is possible for the attacker to perform watermark embedding thus producing a modified watermarked video. On receiving the modified video, the recipient recomputes the watermark and compares this with the watermark extracted from the video. They will match because the embedded watermark and recomputed watermark use the same hash function based watermark generation and the same input i.e. the modified video. Our cryptanalysis strategy applies to any watermarking based video authentication scheme where the watermark and embedding are not functions of secrets. As countermeasure, the functions should be designed so that only legitimate parties can perform them. We present two improved schemes that solve this problem based on private key signing functions and message authentication functions respectively.
Optimal Watermark Embedding and Detection Strategies Under Limited Detection Resources