On the Security of Multiple Encryption or CCA-security+CCA-security=CCA-security?

In a practical system, a message is often encrypted more than once by different encryptions, here called multiple encryption, to enhance its security. Additionally, new features may be achieved by multiple encrypting a message, such as the key-insulated cryptosystems and anonymous channels. Intuitively, a multiple encryption should remain “secure”, whenever there is one component cipher unbreakable in it. In NESSIE’s latest Portfolio of recommended cryptographic primitives (Feb. 2003), it is suggested to use multiple encryption with component ciphers based on different assumptions to acquire long term security. However, in this paper we show this needs careful discussion, especially, this may not be true according to adaptive chosen ciphertext attack (CCA), even with all component ciphers CCA-secure. We define an extended model of (standard) CCA called chosen ciphertext attack for multiple encryption (ME-CCA) emulating partial breaking of assumptions, and give constructions of multiple encryption satisfying ME-CCA-security. We further relax CCA by introducing weakME-CCA (ME-wCCA) and study the relations among these definitions, proving ME-wCCA-security can be acquired by combining IND-CCA-secure component ciphers together. We then apply these results to key-insulated cryptosystem.