💻 computer science

A Robust Fractional Fourier-Domain Image Encryption Framework Integrating Hessenberg Decomposition and Hybrid Chaotic Maps

This paper proposes a robust and efficient image encryption framework that integrates Fractional Fourier Transform, Hessenberg decomposition, and hybrid chaotic maps within a three-stage confusion-diffusion process to achieve high security against various attacks while maintaining low computational complexity.

Original authors: Arabind Kumar

Published 2026-10-21

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Arabind Kumar

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

In the digital age, images travel across open networks every second, carrying everything from medical scans to personal memories. Because these channels are inherently insecure, protecting the data within them is a constant battle. Traditional locks, designed for text, often struggle with pictures because images contain massive amounts of repeating patterns and tightly linked details. To secure them, scientists have long looked toward the realm of frequency, where an image is not seen as a collection of pixels but as a complex wave of energy. By shifting an image into this frequency domain, it becomes possible to scramble its structure in ways that are invisible to the naked eye but mathematically precise. However, simply shifting the image is not enough; the system must also generate a unique, unpredictable key for every single use, ensuring that even a tiny change in the original picture results in a completely different locked version.

A researcher has now proposed a new framework that weaves together three distinct mathematical tools to create a more robust shield for digital images. Their approach begins by transforming the image into a fractional frequency domain, a generalized version of the standard frequency shift that introduces an extra layer of complexity. Once the image is in this new state, the researcher applies a technique called Hessenberg decomposition. This process acts like a filter that simplifies the complex wave patterns of the image, stripping away redundancy while keeping the essential energy concentrated in a compact form. This step not only speeds up the processing but also prepares the data for the final, most critical stage: the application of a hybrid chaotic system.

The core innovation lies in how this system generates the keys used to scramble the image. Instead of relying on a single, simple pattern, the researcher combined two different chaotic maps—one based on a logistic-sine-cosine structure and another on a tangent-delay ellipse reflection. These two systems work together to produce a stream of numbers that is highly unpredictable and sensitive to the slightest change. Before this stream is applied, the researcher feeds the original image into a hash function, a digital fingerprinting tool. This ensures that the key is tied directly to the specific image being encrypted. If even a single pixel in the original picture changes, the entire key changes, making the system incredibly sensitive and resistant to attacks where a hacker tries to guess the key by analyzing patterns.

The results of this new method are striking. When the researcher tested their framework on standard images, the encrypted versions showed no discernible patterns, with pixel values distributed as randomly as a perfect roll of dice. The statistical measures of randomness were nearly ideal, reaching a value of 7.9998 out of a maximum of 8, indicating that the encrypted image reveals almost no information about the original. Furthermore, the system proved highly resilient against attempts to break it. Even when the encrypted image was subjected to noise or had parts of it cut away, the original picture could still be recovered with high clarity, retaining its structural integrity. The researcher found that the system could withstand significant data loss, with the decrypted image remaining recognizable even when a quarter of the encrypted data was missing.

Security against different types of attacks was also rigorously tested. The researcher demonstrated that changing just one bit in the original image caused the entire encrypted result to change completely, a property known as the avalanche effect. This makes it nearly impossible for an attacker to deduce the key by comparing two slightly different encrypted images. The key space, or the total number of possible keys available, was calculated to be larger than two to the power of 512, a number so vast that a brute-force attack would be computationally infeasible with current technology. The entire process, while complex, operates with an efficiency that allows for real-time application, processing images in a fraction of a second on standard hardware.

This work represents a significant step forward in how digital images are protected. By integrating a frequency shift, a matrix simplification technique, and a dual chaotic engine, the researcher has created a system that is not only secure but also efficient and robust against the kinds of data corruption that occur in real-world transmission. The framework successfully bridges the gap between theoretical mathematical elegance and practical security needs, offering a solution that is both mathematically sound and capable of withstanding the rigors of modern digital communication. As multimedia data continues to grow in volume and sensitivity, such integrated approaches provide a necessary foundation for keeping visual information safe in an increasingly open world.

Technical Summary: A Robust Fractional Fourier-Domain Image Encryption Framework Integrating Hessenberg Decomposition and Hybrid Chaotic Maps

Problem Statement
The rapid proliferation of multimedia data transmission over insecure networks has created critical challenges for information security, particularly regarding the protection of sensitive images in fields such as medical diagnosis, military reconnaissance, and commercial transactions. Conventional encryption algorithms (e.g., AES, DES) are often inefficient for images due to the high redundancy, large data volume, and strong correlation between adjacent pixels inherent in image data. Furthermore, existing transform-domain encryption schemes, particularly those based on the Fractional Fourier Transform (FrFT), often suffer from linearity vulnerabilities that make them susceptible to chosen-plaintext attacks. Additionally, while matrix decomposition methods like Hessenberg decomposition have shown promise in reducing computational burdens, their systematic integration with hybrid chaotic systems for spatial-domain diffusion remains under-explored. There is a distinct lack of frameworks that synergistically combine fractional spectral geometry, matrix factorization, and dual-map chaos to achieve high security, low computational complexity, and robustness against transmission errors.

Methodology
The paper proposes a three-stage confusion–diffusion encryption framework that integrates the Fractional Fourier Transform (FrFT), Hessenberg decomposition, and a hybrid chaotic system.

Transform and Decomposition (Stage 1):
- The plaintext image is first transformed into the fractional Fourier domain using the FrFT, where the fractional order $\alpha$ serves as an auxiliary cryptographic key.
- The resulting complex-valued FrFT spectrum undergoes Hessenberg decomposition ( $A = QHQ^*$ ). This process reduces the matrix to an upper Hessenberg form ( $H$ ) and a unitary matrix ( $Q$ ). This step concentrates spectral energy into fewer coefficients, simplifying the spectrum and reducing computational requirements while retaining essential spectral information.
Hybrid Chaotic Keystream Generation (Stage 2):
- A hybrid chaotic engine combines the 2D Logistic-Sine-Cosine Map (2D-LSCM) and the Tangent Delay Ellipse Reflecting Cavity Map System (TD-ERCS).
- To ensure plaintext sensitivity and resistance to chosen-plaintext attacks, a SHA-256 hash of the plaintext image is computed. This hash is used to perturb the initial conditions of the chaotic maps, creating a plaintext-dependent key schedule.
- The combined chaotic sequences generate three distinct keystreams: one for bit-level permutation ( $K_{perm}$ ), one for dynamic S-box substitution ( $K_{sub}$ ), and one for bidirectional diffusion ( $K_{diff}$ ).
Confusion–Diffusion Encryption (Stage 3):
- Confusion: The image is converted to a binary stream, which undergoes bit-level permutation based on $K_{perm}$ .
- Substitution: A dynamic S-box, generated via Fisher-Yates shuffle using $K_{sub}$ , performs substitution on the permuted pixels.
- Diffusion: A bidirectional diffusion process (forward and backward) using $K_{diff}$ and XOR operations ensures that changes in a single pixel affect the entire ciphertext.

The decryption algorithm reverses these steps, utilizing the inverse FrFT, inverse Hessenberg reconstruction, and the inverse chaotic operations.

Key Contributions
The paper identifies four primary contributions:

Novel Tri-Domain Architecture: It proposes a unique encryption architecture combining fractional spectral geometry (FrFT), matrix factorization (Hessenberg decomposition), and dual-map chaos (2D-LSCM and TD-ERCS).
Computational Efficiency via Decomposition: It demonstrates that applying Hessenberg decomposition to FrFT spectra provides significant computational benefits by creating a compact intermediate representation without sacrificing security features.
Enhanced Key Diversity and Sensitivity: It introduces a hybrid chaotic key generation scheme that achieves high key diversity and ensures one-time-pad-like sensitivity to plaintext changes through SHA-256 hash perturbation.
Comprehensive Security Validation: The work provides extensive security proofs and experimental results, validating the system against statistical, differential, chosen-plaintext, and brute-force attacks, as well as noise and occlusion.

Experimental Results
The framework was evaluated using standard test images (Lena, Baboon, Peppers, Cameraman, Airplane) on a MATLAB platform. Key performance metrics include:

Key Space: The key space is calculated to be larger than $2^{512}$ (considering continuous parameters and the SHA-256 hash), far exceeding the $2^{128}$ threshold required to resist brute-force attacks.
Information Entropy: The ciphertexts achieved entropy values between 7.9997 and 7.9999, approaching the ideal value of 8.0, indicating negligible statistical redundancy.
Correlation: Adjacent pixel correlation coefficients were reduced to near zero (within $\pm 0.003$ ), effectively eliminating the correlation present in natural images.
Differential Attack Resistance: The Number of Pixels Change Rate (NPCR) exceeded 99.61%, and the Unified Average Changing Intensity (UACI) was approximately 33.46%, meeting ideal standards for resisting differential attacks.
Robustness: The scheme demonstrated resilience against noise and occlusion. Under 10% salt-and-pepper noise, the decrypted image achieved an SSIM of 0.7836, while under 25% cropping (1/4 area), the SSIM was 0.6915, with the decrypted images retaining recognizable semantic features.
Computational Complexity: The total complexity is dominated by the Hessenberg decomposition ( $O(n^3)$ ) and FrFT ( $O(N \log N)$ ). For a 256×256 image, the total encryption time was approximately 66.4 ms, suggesting suitability for real-time applications.

Significance and Claims
The paper claims that the proposed framework sets a new benchmark in transform-domain chaotic image encryption. By uniquely integrating fractional spectral geometry, matrix factorization, and dual-map chaos, the system overcomes the limitations of traditional schemes, such as linearity vulnerabilities and lack of plaintext sensitivity. The author asserts that the framework offers a viable, efficient, and secure solution for image communication, particularly in environments requiring robustness against transmission errors and various cryptographic attacks. The work suggests that future research could extend this framework to color video, embedded platforms (FPGA/IoT), and quantum-resistant primitives, but the current study focuses on validating the grayscale image encryption performance.

More like this