A Study on the Collusion Security of LUT-Based Client-Side Watermark Embedding

Recently, Celik et al. proposed a lookup-table (LUT)-based embedding mechanism for spread-spectrum watermarks, in which the content distribution server sends an encrypted content to the client, then the client uses his or her personalized decryption LUT to decrypt the received content, meanwhile embeds his or her personalized watermark into the content. They also provided a brief analysis on the security of the LUT-based embedding in terms of the collusion attack on watermarked contents. However, we find that the LUT-based embedding is vulnerable to not only the collusion attack on watermarked contents but the collusion attack on decryption LUTs as well, due to the fact all clients share the same long-term encryption LUT. In this paper, we present a theoretical analysis on the collusion security of the LUT-based embedding mechanism. The analysis shows that the collusion attack on decryption LUTs is more effective than the collusion attack on watermarked contents. Based on our analysis, the content distribution system proposed by Celik et al. can only be used for package sale, which limits its applications. In order to extend the applications, we suggest that the encryption and decryption LUTs of the LUT-based client-side embedding should be set as short-term keys instead of long-term keys. Finally, simulations are carried out to illustrate and validate our theoretical analysis.