Binary fingerprinting codes (Q2486878)

Fingerprinting is the process whereby codewords from a code \(C\) are embedded into otherwise identical digital objects, thus making them unique. These objects are distributed among a number of users. Users working together to create illegal copies are assumed to be unable to create any changes in the objects, other than in places where the objects differ. Such users, called pirates, can construct a new object, but if there are at most \(c\) pirates, a \(c\)-IPP (identifiable parent property) code allows the distributor to trace at least one of the pirates taking part in constructing it. However, to be able to construct a \(c\)-IPP code requires that the alphabet size, \(q\), be strictly greater than the maximum number of pirates, \(c\). For cases where this is not possible, fingerprinting codes with an alphabet size not greater than the maximum number of pirates (\(q \leq c\)), must be employed. This paper considers the case \(q=c=2\). Three binary fingerprinting code classes with properties similar to codes with the identifiable parent property are proposed. In order to compare such codes a new combinatorial quality measure is introduced. In the case of two cooperating pirates, the measure is derived for the proposed codes, upper and lower bounds are constructed and the results of a computer search for good codes according to the quality measure are presented. Some properties of the quality measure are also derived.