Data security in a fixed-model arithmetic coding compression algorithm

Abstract

The data security provided by a fixed-model arithmetic coding compression algorithm is investigated. The “key” to this proposed encryption system comprises the source symbol table and associated frequencies (the model). A chosen plaintext attack is described which allows determination of the key. The major weakness of the algorithm is the generation of non-random output. Repeating binary strings are produced upon input of long strings of one source symbol. This feature can be used to determine the ordering of the symbols in the source symbol table and their associated frequencies. The key is thus discovered, allowing an attacker to decipher intercepted messages. Security may be improved by choosing random symbol ordering and random frequencies. This adversely affects the performance and compression characteristics, however. The algorithm is not recommended for encryption purposes.