A research project analyzing AES applied to digital image encryption - evaluating encryption effectiveness using cryptographic and statistical metrics (NPCR, UACI, entropy, histogram uniformity, pixel correlation) and examining hybrid AES approaches that integrate chaotic systems, dynamic S-boxes, and pixel scrambling.
đź“„ Full Research Paper (PDF)
Role: Team Lead & Primary Research Contributor
- Analyze the AES algorithm and its internal round structure
- Examine how AES is applied to image encryption workflows
- Evaluate encryption strength using statistical and differential analysis metrics
- Investigate hybrid AES approaches that improve multimedia security
- Identify strengths and limitations of AES in modern multimedia environments
AES is a symmetric block cipher standardized by NIST in 2001, operating on 128-bit data blocks with key sizes of 128, 192, or 256 bits.
Each encryption round applies four transformations designed to maximize confusion and diffusion:
| Operation | Function |
|---|---|
| SubBytes | Nonlinear substitution using the AES S-box |
| ShiftRows | Row-wise permutation to disperse byte positions |
| MixColumns | Matrix transformation to strengthen diffusion |
| AddRoundKey | XOR with round key derived from the original key |
These operations collectively transform structured image data into output that is statistically indistinguishable from random noise.
Raw Image
|
v
Convert pixels to byte stream
|
v
Divide into 128-bit blocks (16 bytes each)
|
v
Apply AES encryption rounds per block
|
v
Encrypted image (visual noise - no recoverable structure)
For RGB images, each color channel (Red, Green, Blue) is processed independently, ensuring full spatial and spectral randomization.
Measures what percentage of pixels change when a single pixel in the original is modified. High NPCR confirms strong diffusion.
- Secure threshold: ~99.6% - 99.8%
Measures the average intensity difference between two cipherimages encrypted from slightly different plaintexts. Tests resistance to differential attacks.
- Ideal value: ~33.33%
Measures randomness in the encrypted image data. Maximum entropy for an 8-bit image is 8.0 - values close to this confirm near-perfect randomness.
- AES-encrypted images typically achieve: 7.997 - 7.999
Original images have non-uniform histograms reflecting pixel patterns. Properly encrypted images produce flat, uniform histograms - no visual structure survives encryption.
| Image State | Adjacent Pixel Correlation |
|---|---|
| Original image | > 0.95 (highly correlated) |
| AES-encrypted image | ~0.00 (effectively decorrelated) |
Near-zero correlation confirms that spatial redundancy has been fully eliminated.
Standard AES performs well on image encryption, but research proposes hybrid approaches to further strengthen security against specialized attacks on structured data.
Chaotic systems generate high-quality pseudo-random sequences for key generation or pixel permutation:
- Logistic map
- Henon map
- Lorenz attractor
Chaotic sequences are highly sensitive to initial conditions, making key reconstruction extremely difficult.
Standard AES uses a fixed S-box. Dynamic S-boxes change the substitution mapping per encryption session, significantly increasing resistance to algebraic and differential cryptanalysis.
Spatial redundancy in images is reduced before AES is applied, weakening statistical attacks:
- Arnold Transform
- Zig-zag scanning
- Random permutation
Scrambling destroys positional correlations between adjacent pixels before the encryption layer is applied.
- AES achieves NPCR and UACI values within the secure thresholds for image encryption
- Entropy values of 7.997 - 7.999 confirm near-maximum randomness in encrypted output
- Pixel correlation drops from >0.95 to ~0.00, eliminating exploitable spatial patterns
- Hybrid approaches combining chaotic maps and dynamic S-boxes provide measurable improvements against specialized multimedia attacks
- AES remains the recommended foundation for secure multimedia encryption, with hybrid enhancements appropriate for high-security applications
- Hybrid methods introduce significant computational overhead - relevant for real-time or resource-constrained environments
- Chaotic parameter management adds key management complexity
- No standardization exists for hybrid AES approaches, requiring case-by-case security validation
- New hybrid techniques require extensive cryptanalytic testing before deployment
Cryptographic Algorithm Analysis Multimedia Security Research AES Internals Statistical Security Metrics Differential Cryptanalysis Entropy Analysis Hybrid Encryption Research Academic Research Writing Team Leadership
Research draws from peer-reviewed publications including IEEE, PLOS ONE, and Scientific Reports on AES-based image encryption and hybrid cryptographic techniques.
This project was developed as part of academic coursework at the University of Houston and expanded for cybersecurity portfolio demonstration.
Author: Durga Sai Sri Ramireddy | MS Cybersecurity, University of Houston
