The security of arithmetic coding encryption schemes

This thesis examines a variety of arithmetic coding encryption schemes proposed in existing literature and the attacks against these schemes. A classi­fication scheme for arithmetic coding encryption schemes is proposed and a new adaptive brute-force attack is described. Under certain conditions, it is shown that an adaptive brute-force attack is capable of recovering the key. The way in which the arithmetic coding encryption scheme is initialized appears to make a significant difference to the security of the system as a whole. Although there are modes of operation which are resistant to the adaptive brute-force attack, the security of arithmetic coding encryption schemes is not assured. The results are not conclusive, but are biased toward insecurity. A periodic model resetting scheme was proposed as a means of improving security. However, empirical results indicate that any improvement in security is offset by an equivalent loss in compression performance. This result negates the benefits of unifying compression and encryption. Currently, there is insufficient evidence to recommend arithmetic cod­ing encryption schemes as an efficient and secure scheme for the unification of compression and encryption. The fact that an apparently small difference between the key initialization schemes can make such a dramatic difference to the security of the system is indication that more research is required. Conse­quently, arithmetic coding encryption schemes are not recommended for secure communication purposes