Protecting cipher block chaining against adaptive chosen plaintext attack

In the literature, several encryption modes of operation based on cipher block chaining (CBC) has been proven to be secure under non-adaptive chosen plaintext attack (CPA-1) in the left-or-right (LOR) or find-then-guess (FTG) security models. However, it was shown by Joux et. al. at Crypto 2002 that if we allow the adversary to perform an adaptive chosen plaintext attack (CPA-2), then CBC, ABC and GEM are susceptible to FTG attacks. In this paper, we propose a new CBC-type encryption called input-output masked CBC (IO-CBC) which can protect against FTG and LOR attacks based on forcing an input collision, protects against Joux's FTG attack under proper implementation, and increases the difficulty of linear and differential cryptanalysis. The efficiency of IO-CBC is comparable to CBC because it does only one additonal encryption when compared with CBC. We also reasoned that the security proof of an IO-CBC variant follows from that of OCB.