Loading...

Image Steganalysis Based on Feature Optimization Using Evolutionary Algorithms

Karandish, Mohammad Ali | 2013

631 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 44479 (05)
  4. University: Sharif University of Technology
  5. Department: Electrical Engineering
  6. Advisor(s): Ghaemmaghami, Shahrokh
  7. Abstract:
  8. Image steganalysis is the technique and art of detecting covert communication via images.Reduction of features dimentionality is an important issue according to accuracy and time complexity. In this thesis, GA (genetic algorithm) and PSO (Particle Swarm Optimization) are used to reduce the dimentionality of JRM, a recently proposed feature set containing 11255 features which looks high dimentional compared to other feature sets which has been reduced by evolutionary algorithms so far. So, inspite of other works done using evolutionary algorithms in this field, we use the class sepearability criterion as fitness function instead of the accuracy of the classifier. Investigating these features, we selected the suitable seperability criterion and considered each of 51 feature subsets as a gene in GA or a bit in PSO method. In this phase, the feature dimentionality was reduced to 7500- 8500 without many changes in error rate and PSO method yielded a little better results than GA method. In the next phase of the project, first, we reduced the features dimentionality using PCA method discarding 0.003 of the variance of the data. Then we implemented GA and PSO to reduce the resulting features dimention which led to 650- 750 features (about %94 reduction rate) accepting about %3 error rate. In this situation, the GA method performed better than PSO in error rate and the number of reduced features
  9. Keywords:
  10. Genetic Algorithm ; Steganalysis ; Image Watermarking ; Feature Reduction ; Particles Swarm Optimization (PSO) ; Class Seperability Criterion

 Digital Object List

 Bookmark

No TOC