SECURITY AND PERFORMANCE EVALUATION OF CRYPTOGRAPHIC HASH FUNCTIONS IN ROBUST IMAGE ENCRYPTION

Abstract

This thesis focuses on the security and performance of cryptographic hash functions applied to image encryption algorithms. With the increasing use of digital images that convey confidential data in industries such as healthcare, military, and cloud computing, selecting a suitable hash function becomes crucial to achieve a balance between security strength and computational efficiency. However, most existing image encryption schemes adopt a single hash function—typically SHA-256 or MD5— without an empirical justification for the choice, and the literature lacks a unified, multicriteria comparison that evaluates hash functions from different algorithmic families under identical encryption conditions. Consequently, no clear, evidence-based guidance exists to help designers select a hash function according to an application's specific security and performance requirements. This thesis addresses that gap by evaluating eight hash functions within a single, fixed encryption algorithm using a common set of benchmark images and a shared set of evaluation criteria, thereby enabling a fair, likefor-like comparison and a practical selection framework. The proposed algorithm encrypts an image by partitioning the input image into byte arrays, generating hashes using an initial key, performing XOR operations on the arrays, and finally shuffling the byte arrays according to the generated hash values. The proposed method was examined by using eight hash functions—SHA-1, SHA-384, SHA-512, MD5, Blake2s, RipeMD320, Whirlpool, and Tiger. Six benchmark test images—Baboon, Lake, Peppers, Lena, Petra, and Minaret—were selected for the evaluation. Seven criteria were used to assess security and performance efficiency: Normalized Mean Absolute Error (NMAE), Peak Signal-to-Noise Ratio (PSNR), Entropy, Correlation Coefficient, Key Size, Key Space, and Hash Rate. The results show that SHA-512 is the most secure hash function, achieving the highest NMAE (72.723%), the lowest pixel correlation (0.00297), the highest entropy, and a key space of 2⁵¹², making it suitable for applications where maximum security is paramount. The Tiger hash function achieved the shortest encryption time (0.372 seconds) despite a key space of 2¹⁹², making it well suited to real-time and resourceconstrained settings. Blake2s and RipeMD320 offer a strong balance of fast encryption and high security. Owing to their vulnerability to collisions, MD5 and SHA-1 should be avoided in new implementations. Histogram analysis revealed that the SHA-family functions and MD5 produce the most uniformly flattened histograms. Overall, the study delivers a systematic, multi-criteria framework for selecting hash functions in image encryption systems.