An audio encryption method based on quaternary logic operations

Audio data has strong correlations between adjacent samples, which is challenging for encryption methods. Current audio encryption methods mostly rely on complex transformations and algorithms to ensure their security, but overly complex algorithms will inevitably affect efficiency. We propose a novel audio encryption method by combining quaternary logic operations with MixColumns operations; this method provides higher security than existing audio encryption methods while having faster encryption and decryption speeds. The fast speed is due to the fact that encryption and decryption processes are mainly symbol replacements, which can be implemented very fast by quaternary logic operations. The high security comes from three aspects: (1) Quaternary logic operations can provide a huge space for encryption rules, because there are at most (4!)⁴ types of quaternary logic operations available for encrypting every 2 bits of data; (2) MixColumns operations can reduce the correlation between adjacent samples in audio data, enhancing the diffusion effect while improving key sensitivity; (3) We also introduce dual encryption to resist statistical attacks. To prove our method's security, we analyze key sensitivity and calculate security indicators including key space, number of sample change rate, unified average changing intensity, and peak signal-to-noise ratio, where we compare our method with other existing audio encryption methods. Results show that the security indicators of our method are superior to most of the compared methods. Our method is also faster than existing audio encryption algorithms based on chaos theory, AES encryption, and lightweight password Ascon. When implemented in C++, the encryption speed of our method can reach an average of about 0.000257 s/KB. When implemented on FPGA, the path delay is measured as 2.750 ns reaching a maximum frequency of 363.6 MHz, which is faster than AES, Ascon and other FPGA-based audio encryption methods. Meanwhile, our method consumes fewer FPGA resources than AES and Ascon.