Get started
Work

Titan Vs NIST. A cross valdation Study

Year: 2026
Type: preprint
Abstract: New cipher-evaluation toolkits are often published without evidence that their verdicts are correct. This paper calibrates TITAN, a 42-test, 14-category toolkit developed in India, against the NIST SP 800-22 statistical test suite. Both toolkits were applied to two ciphers: AES-256, the accepted reference, and Atom-Ocean v2, a 64-bit, 8-round Feistel cipher with a 256-bit key proposed by the author. A deliberately weak repeating-key XOR cipher served as a negative control.On the property both toolkits measure, statistical randomness, they agreed in all four ciphertoolkit cases. Under NIST, AES-256 passed 44 of 44 statistics across four independent runs (100%) and Atom-Ocean v2 passed 43 of 44 (97.7%); the single miss lay within chance. The weak control passed 0 of 11. Across 118 validated measurements (NIST and the corrected TITAN v3.1), AES-256 scored 98.3% and Atom-Ocean v2 97.5%, a difference that is not statistically meaningful.Calibration with AES also exposed four defects in TITAN's non-randomness categories: an ECB test defeated by padding, a diffusion test that depends on message length, a byte-level threshold set for bit-level data, and fixed thresholds that give unstable verdicts. The same cipher received grade C under TITAN v2 and grade A under v3. Further checks found two more: a serial test that misapplies the chi-square to overlapping patterns, and timing tests that flag even constant-time AES when run in Python. After seven corrections (TITAN v3.1), three repeated runs gave identical grades: AES-256 and Atom-Ocean v2 both scored A+ in CTR mode and B in ECB mode, failing only the two ECB-mode tests that any ECB cipher should fail, while the weak XOR and Caesar controls scored D and F.The contribution to cryptology is a reproducible method for validating any cipher-evaluation toolkit: agreement with NIST where the two overlap, and known-answer calibration with AES and a weak control where they do not. Passing these tests shows statistical soundness, not proof of Cryptography security
DOI: 10.5281/zenodo.24088564
Language: English
Open Access status: green
OpenAlex ID: W7225612586
Cites:
Cited by:
Related to: