Privacy data protection scheme using memristive hyperchaos and multi-scale block compressive sensing

Abstract

Recently, most meaningful image encryption schemes incorporated with various compressive sensing models have been developed to achieve dual protection of private image data and appearance. However, there exist performance constraints in these schemes in terms of anti-chosen-plaintext attack capability and key management. Aiming at the above issues, a new visually secure image encryption scheme is proposed using multi-scale block compressive sensing (MSB-CS) model and asymmetric integer wavelet transform (IWT) embedding. In this scheme, a memristor model with locally active tristable is first introduced into the oscillator to construct a new 5D memristive hyperchaotic system to generate cipher flows. Then, the non-linear MSB-CS model is designed to compress sparsely-represented plaintext coefficients. After a series of encryption operations, secret image without semantic features is asymmetrically embedded into the same-scale non-secret-involved carrier image. Additionally, both communicating parties, Alice and Bob, acquire shared secret key through the key sharing protocol based on matrix factorization problem. Finally, simulation experiments and comprehensive analysis indicate that the 5D memristive system has complicated hyperchaotic behaviours. In the meantime, the designed encryption scheme possesses better the anti-differential attack capability and reconstruction performance than other recently proposed schemes.