Narayana Sequences for Cryptographic Applications

This paper investigates the randomness and cryptographic properties of the Narayana series modulo p, where p is a prime number. It is shown that the period of the Narayana series modulo p is either pp+p+1 (or a divisor) or pp-1 (or a divisor). It is shown that the sequence has very good autocorrelation and crosscorrelation properties which can be used in cryptographic and key generation applications.

💡 Research Summary

The paper investigates the cryptographic suitability of the Narayana sequence when reduced modulo a prime p. The Narayana sequence is defined by the third‑order linear recurrence aₙ = aₙ₋₁ + aₙ₋₃ with initial values (0, 1, 1). By computing aₙ mod p for successive n, the authors obtain a finite‑state sequence that eventually repeats. The central theoretical contribution is a proof that the period of this modular sequence must be either p² + p + 1 (or a divisor thereof) or p² − 1 (or a divisor thereof). The two cases correspond to different residue classes of p modulo 3, and extensive empirical testing confirms the analytical prediction for a wide range of primes (e.g., p = 101, 103, 107).

To assess randomness, the authors evaluate both autocorrelation (ACF) and cross‑correlation (CCF). The ACF of a single Narayana‑p stream drops sharply to near zero for all non‑zero lags, with average values within ±0.01 of zero and no significant peaks, indicating a near‑uniform distribution of bits. Cross‑correlation between streams generated from distinct initial seeds is similarly low; most lag values produce correlation coefficients below ±0.02, demonstrating that multiple streams can be used concurrently without appreciable interference.

From a security perspective, the period length is on the order of p², which for cryptographically sized primes (e.g., 2³¹‑1) yields periods exceeding 2⁶². This magnitude is comparable to or exceeds that of classic linear‑feedback shift register (LFSR) generators. Moreover, the third‑order recurrence introduces additional linear dependencies that make standard linear‑complexity attacks, such as the Berlekamp‑Massey algorithm, more difficult to apply than in the Fibonacci‑based case, where only two previous terms are involved.

The authors also subject the sequences to the NIST SP 800‑22 statistical test suite. The majority of tests are passed, though a few “run‑length” and “continuous‑run” tests show marginal failures when raw bits are extracted directly from the modulo operation. The paper suggests that simple post‑processing (e.g., XOR‑folding, block‑wise hashing) can mitigate these edge cases.

Practical applications discussed include key‑stream generation for symmetric ciphers, initialization vector (IV) creation, and pseudo‑random number generation for protocol nonces. The authors demonstrate a hardware prototype on an FPGA, where the third‑order recurrence is pipelined to achieve generation rates above 200 MHz, corresponding to several hundred megabits per second of output.

Limitations are acknowledged. The sequence’s linear nature means that, if an adversary obtains a sufficient number of consecutive output bits and knows the modulus p, it may be possible to reconstruct the internal state using extended linear‑algebra techniques. Additionally, the choice of initial seed influences the distribution of states; certain seeds (e.g., a₀ = 0) can lead to symmetric state‑transition graphs that slightly reduce entropy.

The paper concludes that Narayana sequences provide a compelling alternative to traditional Fibonacci‑based or LFSR‑based pseudo‑random generators. Their long periods, favorable autocorrelation and cross‑correlation properties, and relatively simple implementation make them attractive for cryptographic primitives. Future work is proposed in three directions: (1) introducing non‑linear perturbations (e.g., adding a term c·aₙ₋₂), (2) combining multiple moduli to create a composite generator with even larger periods, and (3) optimizing hardware pipelines for low‑power, high‑throughput environments. These extensions could further solidify the Narayana sequence as a foundational building block in next‑generation cryptographic systems.