Analysis of a Buyer-Seller Watermarking Protocol for Trustworthy Purchasing of Digital Contents

In ubiquitous environments where human users get to access diverse kinds of (often multimedia enabled) services irrespective of where they are, the issue of security is a major concern. Security in this setting encompasses both in the interest of the human users as well as their information and objects that they own. A typical kind of transaction interaction among users and/or machines in these environments is that of exchanging digital objects via purchases and/or ownership transfers, e.g. someone buying a song from iTunes via his iPhone, or downloading either bought or rented movies onto a portable DVD player. Here, there is a need to provide trustworthy protection of the rights of both parties; i.e. the seller's copyright needs to be protected against piracy, while on the other hand it has been highlighted in literature the need to protect innocent buyers from being framed. Indeed, if either party cannot be assured that his rights are protected when he is involved in transactions within such environments, he would shy away and instead prefer for instance the more conventional non-digital means of buying and selling. And therefore without active participation from human users and object owners it is difficult to fully kick off the actual realization of intelligent environments. Zhang et al. recently proposed a buyer-seller watermarking protocol without a trusted third party based on secret sharing. While it is a nice idea to eliminate the need of a trusted third party by distributing secret shares between the buyer and the seller such that neither party has knowledge of the fingerprint embedded in a content, we show that it is possible for a buyer to remove his part of the fingerprint from the content he bought. This directly disproves the piracy tracing property claimed by the protocol. In fact, since piracy tracing is one of the earliest security applications of watermarking schemes, it raises doubts as to the soundness of the design of this protocol.