DOI: 10.35377/saucis...1926017 ISSN: 2636-8129

Matrix M: Reducing Ciphertext Repetition and Structural Leakage in Block Ciphers

Kerim Yıldırım
Show PDF Cite
This study aims to reduce repetitions and structural traces on the ciphertext side of existing block ciphers. Within this scope, the message-transformation layer called Matrix M was designed not as a new block cipher or a new mode of operation, but as a deterministic and reversible transformation layer added before existing ciphers such as AES, PRESENT, CLEFIA, MISTY1, and KASUMI. The evaluation was conducted as an ECB stress test, where block equality is intentionally visible, and repetition leakage can be measured directly. To analyze whether Matrix M produces the intended effect, a benchmark was designed and conducted over 24 full-profile datasets. In the analysis, performance, output expansion, distribution, diffusion, and structural leakage indicators were measured together on the same datasets. The results show that Matrix M reduces the ratio of repeating ciphertext blocks and the dominant block density in all examined ciphers, while increasing the ratio of unique blocks. The findings also reveal that it weakens the direct reflection of repetition patterns in the plaintext onto the ciphertext. These effects are interpreted as changes in the block-equality classes presented to the ECB mode rather than as cipher-internal diffusion improvements. The evaluation also shows the expected cost of this transformation layer, including about 25% wire expansion and, in the AES-ECB case, an approximately 12x increase in total processing time. The main contribution of Matrix M can be summarized as a blurring of structural traces on the ciphertext side and reducing repetition-based leakage under the stated empirical observer model.

More from our Archive