On the Development of Computational Intelligence Steganalysis for Lossless Steganographic Techniques

Reversible data hiding is the technique that guarantees cover-media will fully restored without any distortion after extracting hidden messages from stego-image. Reversible data hiding is suitable for applications in specific domains, such as medical, military, forensic and fine art images. Due to the rapid evolution in digital technology, the application of steganography in various fields continues growing. However, the concern about illegal usage of steganographic techniques for malicious users is also increasing. Stegnalysis is the art of breaking steganographic schemes. The main objectives of steganalysis are filtering out the secret communication and detecting the presence of hidden messages in various medium. Usually, steganalysis has been classified as target-based steganalysis and universal steganalysis methods. Target-based steganalysis can reveal secret message with the knowledge of the steganography algorithm. Universal steganalysis can detect the secret message independent of the steganography algorithm. This project is aims to develop steganalysis schemes against to the novel reversible steganographic methods including difference expansion, histogram shifting and multiple predictors. This project will borrow the ideas from computational intelligence techniques, such as fuzzy clustering, self-organizing feature map networks (SOFM), adaptive resonance theory (ART) to extend the ability of the proposed steganalysis system. The goal of this project is to integrated statistical theories and computational intelligence techniques developing high performance steganalysis schemes with the reliability, flexibility, and practicality for unknown steganographic methods.

Project ID:PB10007-0172
External Project ID:NSC100-2221-E182-055