Region of interest encryption for color images based on a hyperchaotic system with three positive Lyapunov exponets

• Published in: Optics and laser technology Vol. 106; pp. 506 - 516
• Main Authors: Xue, Han-wen, Du, Juan, Li, Shou-liang, Ma, Wei-jiao
• Format: Journal Article
• Language: English
• Published: Kidlington Elsevier Ltd 01.10.2018; Elsevier BV
• Subjects: Algorithms; Color image encryption; Color imagery; Columns (process); Computer vision; Cryptography; Cybersecurity; Data encryption; Diffusion; Encryption; Entropy (Information theory); Histograms; Hyperchaos; Liapunov exponents; Pixels; ROI; Sensitivity analysis; Three positive Lyapunov exponents; Wide parameter range; Hyperchaos; Three positive Lyapunov exponents; Color image encryption; ROI; Wide parameter range
• ISSN: 0030-3992; 1879-2545
• DOI: 10.1016/j.optlastec.2018.04.030

•A hyperchaotic encryption method is proposed for region of interest in color image.•A hyperchaotic system with 3 positive Lyapunov exponents is proposed to encryption.•This system is in hyperchaotic state with wide parameter region.•Hierarchical keys are adopted, and the encryption method is related to plaintext.•The experimental results show the method is robust and has excellent performance. We propose a hyperchaotic encryption algorithm for the region of interest (ROI) in a color image. The ROI for the human face is first identified based on the Open Source Computer Vision library by using a Gaussian mixture model. It is then encrypted using a novel hyperchaotic system, which has three positive Lyapunov exponents. The system can be in a hyperchaotic state with the parameter changing over a wide region, which is proved by Lyapunov exponent spectrum calculations. Furthermore, hierarchical keys are adopted to ensure the security of the encrypted sequences generated. In the encryption process, crossed image row–column scrambling and pixel value diffusion are experienced. The hyperchaotic sequences required for the encryption relate not only to the keys but also to the plaintext images. Consequently, the algorithm is effective against plaintext and ciphertext attacks. The security performance is also verified in detail with respect to histograms, correlations, information entropy, key space, key sensitivity analysis, number of pixel change rate, unified average changing intensity, and known and chosen attacks. The results indicate that the proposed algorithm has a large key space and a low correlation between adjacent pixels, hence, it can also be usefully employed to resist brute-force and statistical attacks.