A Robust Image Encryption Algorithm Based on a 32-bit Chaotic System

Hao Li, Lianbing Deng, Zhaoquan GuIEEE Access 2020 · OpenReview

Abstract

In recent years, image encryption systems based on chaotic maps have received considerable attention. For the security of encryption algorithms, it is vital to ensure that chaotic systems produce high-quality pseudo-random sequences. However, when a chaotic system is deployed on a limited-precision platform, its random performance often deteriorates—a phenomenon known as random degeneration. We prove that even at 256-bit precision, piecewise linear chaotic maps can fail to achieve desired randomness. To address this, we propose a 32-bit positive integer random sequence generator, 32-bit PL_PWLCM&LM, which utilizes a parallel structure of two 32-bit PL_PWLCMs and one 32-bit logistic map. Our encryption system adopts a 4-pixel permutation unit and row/column-based diffusion. Security and performance analyses demonstrate that the proposed scheme is secure, robust, and fast.

The Challenge of Finite Precision

Chaotic maps are ideal for encryption due to their extreme sensitivity to initial conditions and unpredictable outputs. However, real-world hardware (32-bit or 64-bit CPUs) forces these theoretical infinite-precision maps into limited state spaces. This leads to "random degeneration," where sequences become periodic or lose statistical randomness.

The authors demonstrate that standard cascading techniques—often used to increase complexity—actually exacerbate this degeneration in low-precision environments. Even at high precisions like 256-bit, certain maps fail standard randomness tests when implemented traditionally.

The 32-bit PL_PWLCM&LM Generator

To overcome these limitations, the researchers developed the 32-bit PL_PWLCM&LM Pseudo-Random Number Generator (PRNG). Key features include:

  1. Parallel Architecture: Instead of cascading, the system runs two Piecewise Linear Chaotic Maps (PWLCM) and one Logistic Map (LM) in parallel. Their outputs are combined via XOR and MOD operations, avoiding the degeneration traps of cascaded systems.
  2. Integer Domain Operations: By performing all calculations in the 32-bit integer domain, the algorithm eliminates expensive floating-point processing. This makes the system "robust," ensuring identical encrypted outputs across different hardware platforms regardless of their floating-point implementations.
Comparison of chaotic properties (Feigenbaum, Lyapunov, and Sensibility) between standard Logistic and the proposed 32-bit PL_Logistic maps, demonstrating the preservation of chaotic behavior in the integer domain.

As shown in the figure above, the 32-bit integer-based (PL) versions of the maps successfully preserve the bifurcation structures (Feigenbaum diagrams), Lyapunov exponents, and initial condition sensitivity of their theoretical counterparts.

Optimized Permutation-Diffusion Network

The encryption process is further optimized through a novel Permutation-Diffusion Network (PDN):

  • 4-Pixel Permutation Unit: Rather than sorting every individual pixel, the algorithm uses a 2x2 block (4 pixels) as the basic unit. This reduces the required volume of random numbers by 75%, significantly boosting processing speed without compromising security.
  • Row and Column Diffusion: The system employs diffusion across entire rows and columns. This ensures that a single-bit change in the original image spreads across the entire ciphertext within just two rounds, providing high sensitivity to plaintext changes.

Performance and Security Analysis

The algorithm was tested using standard benchmarks from the USC-SIPI image database (e.g., Clock, Man, and various color images). Key results include:

  • Randomness: The generator passed all 15 NIST STS tests and the TestU01 suite at 32-bit precision.
  • Sensitivity: The system is highly sensitive to the 156-bit key; a single-bit change results in a completely different, noise-like ciphertext (NPCR > 99.6%).
  • Speed: On a standard i5 CPU, the algorithm achieved an average encryption speed of 1.827 Mbps, outperforming several recent state-of-the-art schemes.
  • Robustness: Decrypted images remain recognizable even when the ciphertext suffers from 1% salt-and-pepper noise or 6.25% (1/16th) data loss.

Conclusion

By shifting from cascaded floating-point maps to parallel integer-domain chaotic systems, this research provides a high-performance solution for secure image encryption on limited-precision hardware. Future work will focus on adapting this architecture for the high-throughput requirements of real-time 4K video encryption.

Summary updated

Abstract

In these years, image encryption systems based on chaotic maps have received considerable attention from researchers. For the security of encryption algorithms, it is important to ensure that chaotic systems can produce high quality pseudo-random sequences. However, when a chaotic system that theoretically has good random performance is deployed in a limited precision platform, its random performance will deteriorate. We prove that even if the precision reaches 256-bit, the piecewise linear chaotic map still fails to achieve the desired random performance. Moreover, in the case of low precision, the cascade structure commonly used in mixed chaotic maps will aggravate the problem of random degeneration. To generate good pseudo-random numbers in low-precision platforms, we propose a positive integer random sequence generator, called 32-bit PL_PWLCM&LM, which comprises two 32-bit PL_PWLCMs and one 32-bit logistic map in parallel. In our proposed image encryption system, we adopt the permutation algorithm with four pixels as a permutation unit and the diffusion algorithm based on rows and columns. The security and performance analyses prove that the proposed scheme is a secure, robust, and fast image encryption algorithm. Source codes are shared on GitHub: https://github.com/cuclihao/Pic_Enc_V3.
BibTeX
@inproceedings{li2020robust,
  author = {Hao Li and Lianbing Deng and Zhaoquan Gu},
  title = {{A Robust Image Encryption Algorithm Based on a 32-bit Chaotic System}},
  booktitle = {IEEE Access},
  year = {2020},
  url = {https://openreview.net/forum?id=Q9-CNsi9PuJ},
}
Figure