Steganography, the ancient art for secretive communications, has revived on the Internet by hiding secret data in completely imperceptible manners, and has created a serious threat due to the covert channel that can be readily exploited for various illegal purposes. Likewise, multimedia tampering, which has been greatly facilitated and proliferated by various multimedia processing tools, is increasingly causing problems concerning the authenticity of digital multimedia data. There is a critical need to develop reliable methods for steganography detection or steganalysis and for forgery detection to serve purposes in national security, law enforcement, and cybercrime fighting.
To detect steganography and forgery on multimedia data, our research goals include discovering the characteristic modification caused by digital multimedia steganography and forgery, developing more accurate and more reliable methods for steganalysis and digital evidence authentication, and developing a complete evaluation procedure for gaining full understanding of the accuracy, reliability, and measurement validity of steganography detection and digital evidence authentication in digital image, audio, and video files. To achieve these goals, the procedures of our research design and methods are conducted as follows:
- Construct a comprehensive and high volume multimedia steganography and forensic database.
- Analyze the bias and variation of each confirmed source by using existing methods; and by developing new methods, improve the quantification of the characteristics and uncertainties of the cover, steganography, and forgery, created by these sources, and provide a more complete evaluation in different circumstances including multimedia type and format, signal complexity, source type, information-hiding/forgery type and modified size, detection method, detection accuracy, the strength and limitation of a certain method in which circumstance.
- Measure detection performance.
- Monitor and improve the steps in the forensic evidence analysis process in digital media by integrating updated methods with the use of data mining and computational intelligence techniques for steganography detection and forgery detection.
We conjecture that data hiding in steganography and manipulation in forgery production change the statistics of original multimedia data, and hence leave the clues of modification. Our study aims to discover the features that may discriminate the manipulations from intactness and analyze different patterns caused by different operations. In this project, we have developed several novel detection algorithms based on feature mining and machine intelligence techniques in detecting steganography, forgery manipulation and relevant operations such as cropping, double compression on multimedia data. Our experimental results validate our hypothesis and indicate that our methods have obtained the detection performances in detecting several types of steganography and forgery on multimedia data within the state-of-the-art. Our study also shows that a complete evaluation of the detection performance of different algorithms should include image/signal complexityin addition to other relevant factors such as hiding ratio or compression ratioas a significant and independent parameter for some detections including JPEG double compression.