Source-linked AI summary
Image Encryption Using Differential Evolution Approach in Frequency Domain
Maaly Awad S Hassan, Ibrahim Soliman I Abuhaiba
TL;DR
The paper addresses the need for secure image encryption in applications where existing methods may be insecure. It proposes keyed Fourier-domain encryption combined with Differential Evolution operations, and reports fully distorted encrypted images with successful visual recovery using the shared key. The method assumes a shared secret key for selecting and modifying components.
Problem
Image encryption is needed to protect image content in storage and transmission, but some existing methods have been reported as insecure.
Method
The scheme applies a keyed 2-D discrete Fourier transform, LFSR-selected crossover, and keyed mutation within the Fourier domain.
Results
The encrypted images are reported as fully distorted or invisible, while decryption with the same secret key successfully restores grayscale, color, and mesh-text images.
Takeaways & Limitations
The proposed Differential Evolution frequency-domain cryptosystem is reported to hide image information effectively and operate on both grayscale and color images.
Takeaways & Limitations
The scheme requires a secret key shared between the encryption and decryption sides, which initializes the LFSR seed and determines selected indices.
Abstract
from arXiv · showhide
This paper presents a new effective method for image encryption which employs magnitude and phase manipulation using Differential Evolution (DE) approach. The novelty of this work lies in deploying the concept of keyed discrete Fourier transform (DFT) followed by DE operations for encryption purpose. To this end, a secret key is shared between both encryption and decryption sides. Firstly two dimensional (2-D) keyed discrete Fourier transform is carried out on the original image to be encrypted. Secondly crossover is performed between two components of the encrypted image, which are selected based on Linear Feedback Shift Register (LFSR) index generator. Similarly, keyed mutation is performed on the real parts of a certain components selected based on LFSR index generator. The LFSR index generator initializes it seed with the shared secret key to ensure the security of the resulting indices. The process shuffles the positions of image pixels. A new image encryption scheme based on the DE approach is developed which is composed with a simple diffusion mechanism. The deciphering process is an invertible process using the same key. The resulting encrypted image is found to be fully distorted, resulting in increasing the robustness of the proposed work. The simulation results validate the proposed image encryption scheme.
1. INTRODUCTION
Image security is increasingly important as images become central to networked information exchange and sensitive applications. The paper motivates a frequency-domain encryption scheme using keyed DFT and Differential Evolution operations.
- Image encryption protects image confidentiality during storage and transmission by transforming content into an unintelligible image without the decryption key.The paper identifies medical imaging, military communications, pay-TV, and confidential video conferences as relevant applications.
- Existing image-encryption methods include spatial-domain and frequency-domain approaches, with some proposed techniques reported as insecure.The paper distinguishes direct pixel manipulation from Fourier-domain modification.
- The proposed scheme combines keyed discrete Fourier transform with Differential Evolution for magnitude and phase manipulation.The Fourier transform can be inverted to return to the original domain without information loss.
2. RELATED WORK
Prior image-encryption research uses digital signatures, SCAN patterns, vector quantization, optical phase masks, chaotic sequences, and related compression or coding techniques. These methods offer different benefits, including authenticity, lossless processing, compact representation, and security, while also carrying limitations such as increased processing time or distortion risk.
- The related-work landscape includes multiple distinct image-encryption strategies rather than a single standard approach.The section headings identify digital signatures, SCAN, vector quantization, optical phase encoding, and chaotic sequences.
- Digital-signature encryption embeds an image signature into an encoded image, providing authenticity verification and three security layers.The method uses an error-control code and adds redundancy, which changes the image dimension.
- SCAN-based processing performs lossless compression and encryption of binary and grayscale images using generated spatial-accessing patterns.Its stated drawback is longer compression-encryption time.
- Vector-quantization encryption decomposes images into vectors and sequentially encodes them before applying traditional cryptosystems.The cited advantages include simple hardware structure and a small required bit rate.
- Optical color-image encryption converts color images to indexed formats and uses two random phase masks to encode them as stationary white noise.Decryption converts the recovered indexed images back to RGB format.
- A chaotic-sequence method encrypts images in the DCT domain and can restore the original image despite distortion from lossy compression and noise.The method generates key-dependent chaotic sequences and uses symbol and transformation matrices.
3. PROPERTIES OF FOURIER TRANSFORMS: Importance of phase and magnitude
A Fourier transform represents an image through magnitude and phase, and both are required for complete reconstruction. Using only one component produces a degraded or unrecognizable restored image.
- Both magnitude and phase functions are necessary to reconstruct an image completely from its Fourier transform.The paper explains that neither component alone preserves sufficient information for full restoration.
- Restoration using magnitude alone produces an unrecognizable image with severe dynamic-range problems.Figure 1 identifies this restoration as the case where φ = 0.
- Restoration using phase alone produces an image that is barely recognizable and severely degraded in quality.Figure 1 identifies this restoration as the case where |A| is constant.
4. THE METHOD OF DIFFERENTIAL EVOLUTION
Differential Evolution uses crossover and mutation to modify candidate vectors. Crossover recombines parts of two vectors, while mutation changes a selected element to create a different vector.
- Differential Evolution has two main stages: crossover and mutation.The two operations alter vectors through recombination and targeted modification.
- Crossover: Crossover divides two selected binary vectors at a crossover point and swaps corresponding parts to create new vectors.In the illustrated operation, the second part of one vector is exchanged with the second part of the other.
- Mutation: Mutation applies a mutation function to a selected element of a vector, producing a changed vector.The example uses a NOT function on the fourth bit of a binary vector.
5. THE PROPOSED ENCRYPTION METHOD
The proposed encryption method combines a keyed two-dimensional Fourier transform with LFSR-selected crossover and mutation operations. These reversible stages manipulate frequency-domain components and use the shared secret key to support encryption and decryption.
- Keyed Fourier transform: The scheme begins with a keyed 2-D discrete Fourier transform of the original image, making inverse transformation dependent on the secret key.The keyed transform is presented as the first encryption level and is illustrated as reversible when the key is used.
- Overall cryptosystem: The encryption pipeline consists of keyed discrete Fourier transform, crossover, and mutation operations, with shared-key regeneration making the crossover and mutation stages invertible.The method uses LFSR-generated indices to shuffle component positions and produce a fully distorted encrypted image.
- Keyed Fourier transform: The keyed Fourier transform modifies the traditional transform with multipliers involving the secret key while preserving the required multiplier-product restriction.The paper places key-dependent multipliers in the transform and inverse-transform expressions.
- DE crossover operation: Crossover combines two selected frequency-domain components by swapping their complex parts, changing the amplitude and phase of the resulting components.The components are selected using indices generated from an LFSR initialized with the shared secret key.
- DE crossover operation: The crossover operation is reversible because both encryption and decryption regenerate the same component indices from the shared secret key.The paper demonstrates reversibility for the crossover stage using the same key-derived indexing process.
- DE mutation operation: Mutation further changes selected frequency-domain components through a keyed function applied to their real parts, with positions updated by an LFSR generator.The procedure initializes row and column indices from the secret key and repeatedly updates them during mutation.
6. EXPERIMENTAL RESULTS AND SECURITY ANALYSIS
Experiments on grayscale, color, and mesh-text images show that the proposed DE cryptosystem produces visually obscured cipherimages and supports decryption with the same secret key. Security analysis examines robustness through key-space, histogram, and correlation tests.
- Experimental results: The method encrypts 256x256 grayscale Cameraman, 447x301 color Forest, and mesh-text images into visually invisible cipherimages while enabling decryption with the same key.The experiments report successful encryption and decryption across grayscale, color, and complicated mesh-text images.
- Security analysis: The proposed cryptosystem is evaluated against cryptanalytic, statistical, and brute-force attacks through key-space, histogram, and key-sensitivity analyses.The security section explicitly identifies these analyses as tests of the proposed method.
- Key-space analysis: 2^256 secret-key combinations are provided, with a 256-bit LFSR seed initialized by the secret key.The reported key space is intended to make brute-force attacks infeasible for practical use.
A. Histograms analysis
Histogram analysis finds that cipherimage intensity distributions differ substantially from those of the corresponding plainimages, reducing visible statistical resemblance.
- Histogram comparison: Cipherimage histograms are significantly different from the corresponding plainimage histograms and bear no statistical resemblance to them.Plainimage histograms contain large spikes, whereas the encrypted-image histograms do not provide clues for the stated statistical attack.
- Histogram comparison: The analysis compares histograms of original and encrypted Cameraman and Monaliza images to assess statistical similarity.The figures present original-image and cipherimage histograms for these examples.
B. Correlation coefficient analysis
Correlation and key-sensitivity analyses show that encryption greatly reduces adjacent-pixel correlation and that small key changes produce unrelated cipherimages or failed decryption.
- Correlation coefficients: The correlation analysis randomly samples 2000 adjacent-pixel pairs in horizontal, vertical, and diagonal directions before calculating correlation coefficients.The coefficients are computed using the stated covariance and correlation formulas.
- Correlation coefficients: Horizontal adjacent-pixel correlation changes from 0.9898 in the plainimage to 0.0303 in the cipherimage.The paper reports negligible cipherimage correlation compared with high plainimage correlation, with similar results for vertical and diagonal directions.
- Key sensitivity analysis: Changing one bit of the encryption key produces a completely different cipherimage, with no correlation reported among images encrypted under slightly different keys.The comparison uses keys differing in the most significant or least significant bit.
- Key sensitivity analysis: Decryption fails completely when an encrypted image is processed with a slightly different secret key.The experiment contrasts decryption with the original hexadecimal key and a key differing by one bit.
- Lossless and opacity: The experiments report lossless recovery and high distortion between original and encrypted images.Decrypted images are described as exactly similar to the originals, while encrypted images are highly distorted.
- Complexity: The reported encryption and decryption operation times are recorded in Table 3 to measure computational complexity.The supplied passage identifies the timing test but does not provide its numerical values.
7. CONCLUSION AND FINAL THOUGHTS
The paper proposes frequency-domain image encryption using DE crossover and mutation on Fourier components, supported by key-space, statistical, and key-sensitivity analyses. It concludes that the method is expected to be useful for real-time image encryption and transmission, with future work extending it to multimedia and compression.
- Contribution: The proposed scheme applies DE crossover and mutation to frequency-domain components, changing amplitude and phase to increase confusion and diffusion in cipherimages.The method is presented as a new image-encryption scheme for digital-image communications.
- Security evaluation: Key-space, statistical, and key-sensitivity analyses are used to demonstrate the security of the proposed image-encryption procedure.These analyses are the paper's stated basis for its security conclusion.
- Conclusion: The proposed method is expected to be useful for real-time image encryption and transmission applications.The conclusion also identifies multimedia data and image compression as future research directions.
Authors
The paper lists Prof. Ibrahim Abuhaiba and Eng. Maaly Hassan as authors with backgrounds in computer engineering and image processing.
- Prof. Ibrahim Abuhaiba is a professor in the Computer Engineering Department at the Islamic University of Gaza.
- He earned philosophy master's and doctoral degrees in Britain focused on document understanding and pattern recognition.
- His research interests include computer vision, image processing, document analysis, pattern recognition, and artificial intelligence.
- Eng. Maaly Hassan is pursuing a master's degree in Computer Engineering at the Islamic University of Gaza.
- Her interests include image processing, multimedia, and mobile ad hoc and sensor networks.