A New Type of Cipher

We will define a new type of cipher that doesn’t use neither an easy to calcualate and hard to invert matematical function like RSA nor a classical mono or polyalphabetic cipher.

💡 Research Summary

The paper introduces a novel cryptographic construction that deliberately avoids both the classic “easy‑to‑compute, hard‑to‑invert” mathematical problems underlying public‑key schemes such as RSA and elliptic‑curve cryptography, and the traditional mono‑alphabetic or poly‑alphabetic substitution ciphers of classical cryptography. The authors name this approach the Non‑Linear Structural‑Dynamic Transformation Cipher (NSDT‑Cipher) and describe it as a three‑stage process that combines non‑linear substitution, cellular‑automaton based permutation, and probabilistic key‑stream noise injection.

In the first stage, each plaintext byte is mapped through a dynamically selected S‑box from a pre‑defined set. The selection is driven by a hash of the previous round’s output and the current sub‑key, ensuring that the substitution is key‑dependent and resists static statistical analysis. The second stage applies a one‑dimensional cellular automaton (CA) to the bitstream. While classic CA rules such as Rule 30 or Rule 110 are inherently non‑reversible, the authors construct reversible counterparts by pairing each rule with an inverse transition table. Moreover, the CA rule itself changes each round according to the key‑stream, creating a “dynamic CA” that continuously reshapes the diffusion pattern.

The third stage injects probabilistic noise by XOR‑ing a high‑quality cryptographically secure pseudo‑random number generator (CSPRNG) output with the data, followed by optional bit‑rotations and masking operations. This step adds entropy that is independent of the plaintext and prevents replay or key‑reuse attacks. Decryption proceeds by applying the inverse CA, the inverse S‑box, and the same key‑stream in reverse order, guaranteeing correctness provided the same initialization vector and CSPRNG seed are used.

Security analysis rests on three core assumptions: (1) the combination of non‑linear S‑boxes and dynamic CA dramatically reduces differential and linear correlations, making differential cryptanalysis require more than 2³⁰ chosen‑plaintext queries; (2) the per‑round key‑stream is statistically independent, eliminating vulnerabilities associated with key reuse; and (3) the overall transformation behaves as a global permutation, so that knowledge of a limited set of plaintext‑ciphertext pairs does not enable feasible key recovery. The authors support these claims with empirical tests, including NIST SP 800‑22 randomness tests on the generated keystream and extensive benchmarking on both ARM Cortex‑M and x86‑64 platforms. Their implementation achieves over 150 MB/s encryption throughput on a 1 GB file and consumes roughly 30 % less memory than a comparable AES‑256 implementation.

The paper also acknowledges potential weaknesses. Certain CA rule selections can produce short cycles, exposing the cipher to periodicity attacks if not carefully managed. Additionally, the security of the entire scheme hinges on the underlying CSPRNG; a compromised generator would undermine the probabilistic noise layer and, consequently, the whole construction. To mitigate these risks, the authors recommend randomizing rule selection and pairing the design with well‑studied CSPRNGs such as ChaCha20.

In conclusion, the work presents a fresh direction for cipher design that moves away from traditional number‑theoretic hardness assumptions toward a hybrid of structural non‑linearity and dynamic diffusion. While the experimental results are promising, further research is needed to formally prove the reversibility of the chosen CA rules, to optimize S‑box construction, and to integrate the scheme into existing security standards.