Exploring a Q-learning-based chaotic naked mole rat algorithm for S-box construction and optimization
This paper introduces a new variant of the metaheuristic algorithm based on the naked mole rat (NMR) algorithm, called the Q-learning naked mole rat algorithm (QL-NMR), for substitution box construction and optimization. Unlike most competing works (which typically integrate a single chaotic map int...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English English |
| Published: |
Springer Science and Business Media Deutschland GmbH
2023
|
| Subjects: | |
| Online Access: | http://umpir.ump.edu.my/id/eprint/40759/1/Exploring%20a%20Q-learning-based%20chaotic%20naked%20mole%20rat%20algorithm.pdf http://umpir.ump.edu.my/id/eprint/40759/2/Exploring%20a%20Q-learning-based%20chaotic%20naked%20mole%20rat%20algorithm%20for%20S-box%20construction%20and%20optimization_ABS.pdf http://umpir.ump.edu.my/id/eprint/40759/ https://doi.org/10.1007/s00521-023-08243-3 https://doi.org/10.1007/s00521-023-08243-3 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | This paper introduces a new variant of the metaheuristic algorithm based on the naked mole rat (NMR) algorithm, called the Q-learning naked mole rat algorithm (QL-NMR), for substitution box construction and optimization. Unlike most competing works (which typically integrate a single chaotic map into a particular metaheuristic algorithm), QL-NMR assembles five chaotic maps (i.e., Chebyshev, logistic, circle, Singer, and sinusoidal) as part of the algorithm itself. Using a Q-learning table, QL-NMR remembers the historical performance of each chaotic map during the S-box construction process allowing just-in-time adaptive selection based on its current performance. Experimental results for 8 × 8 S-box generation demonstrate that the proposed QL-NMR gives competitive performance against other existing works, particularly in terms of nonlinearity and strict avalanche criteria. To further demonstrate the effectiveness of our proposed work, we have subjected the QL-NMR for image segmentation using multilevel thresholding. The results confirm that QL-NMR gives better performance than its predecessor NMR. Finally, QL-NMR S-box also outperformed NMR S-box in image encryption. |
|---|