Steganalysis Research Based On Artificial Immune System

Steganalysis is a significant technique in keeping steganography within limits, ensuring both national and military information security and social stability. In the face of so abundant network digital media, steganalysis has increasingly required for higher detection accuracy and intelligence. To enhance stego-detectors' intelligence such as self organizing and self learning, AIS inherent merits of immune recognition and immune learning can be used in steganalysis. This thesis aims to design a new effective, universal and intelligent steganalysis method based on AIS. The main contribution and innovation are summarized as follows:1. The basic idea and whole frame of steganalysis based on AIS are proposed. By analyzing the commonness between steganalysis and AIS, steganalysis based on AIS considered is laid out and divided into two parts: extracting classified features for steganalysis and designing the steganalysis detectors based on AIS.2. A new effective and universal high order statistic related feature (RR-Farid) is extracted. By theoretically analyzing the classical Farid high order statistic linear prediction model and revising its inconsistent relativity area, an R-Farid high order statistic model firstly proposed to enhance feature's effectivity. Then, for lower relying on specific stego-domain or carrier sample, a rational re-embed operation and the basic hypothesis of its random information are analyzed. Finally, as follow them, RR-Farid is extracted as feature for steganalysis based on AIS.3. The stego-detectors based on AIS are designed and realized. Closely integrating with bionic mechanisms in AIS and the data characteristic of RR-Farid, the primary process of steganalysis based on AIS is firstly designed and formalized. Then, combined with variable deteted radiu, space cover estimation and syncretism technique, DGA_BAIS (Detectors Generation Algorithm Based AIS) and DDOU_BAIS (Detectors Detecting and Dynamic Optimized UpDate Algorithm Based AIS) are brought forward. Experimental results show, the stego-detectors based on AIS not only can be optimized and matured by themselves to improve detection rate, which better than conventional stego-classifier like FLD and SVM, but also able to effectively discriminate the clean and dirty carriers whose feature space been parted by discrete boundaries, even sense some unknown steganographies.Finally, the research topics are summarized and prospected.