An Ambiguous Coding Scheme for Selective Encryption of High Entropy Volumes

This study concentrates on the security of high-entropy volumes, where entropy-encoded multimedia files or compressed text sequences are the most typical sources. We consider a system in which the cost of encryption is hefty in terms of some metric (e.g., time, memory, energy, or bandwidth), and thus, creates a bottleneck. With the aim of reducing the encryption cost on such a system, we propose a data coding scheme to achieve the data security by encrypting significantly less data than the original size without sacrifice in secrecy. The main idea of the proposed technique is to represent the input sequence by not uniquely-decodable codewords. The proposed coding scheme splits a given input into two partitions as the payload, which consists of the ambiguous codeword sequence, and the disambiguation information, which is the necessary knowledge to properly decode the payload. Under the assumed condition that the input data is the output of an entropy-encoder, and thus, on ideal case independently and identically distributed, the payload occupies ~~ (d-2)/d, and the disambiguation information takes ~~ 2/d of the encoded stream, where d>2 denotes a chosen parameter typically between 6 to 20. We propose to encrypt the payload and keep the disambiguation information in plain to reduce the amount of data to be encrypted, where recursive representation of the payload with the proposed coding can decrease the to-be-encrypted volume further. When 2 * 2^d <= n <= tau * d * 2^d, for tau = (d-1.44)/2, we show that the contraction of the possible message space 2^n due to the public disambiguation information is accommodated by keeping the codeword set secret. We discuss possible applications of the proposed scheme in practice.