Morphological pattern spectrum and block cipher processing based image-manipulation detection

This paper presents a morphological pattern spectrum-based image manipulation detection method with a highly parallel two-dimensional cellular automata architecture (CAM). The novel method consists of using a mathematical morphological-based algorithm to extract pictorial feature information from an original digital image and a block-cipher algorithm to protect the extracted-pictorial information. This will be useful in situations where it is important to find evidence of specific events such as in the investigations of crimes. The morphological pattern spectrum was implemented with a CAM instruction-set program and tested with an evaluation system. The CAM processor achieves highly parallel processing with low power consumption and is thus effective for mobile product applications. Analyses of manipulated-images indicated that the proposed detection method was able to clearly identify differences from the original image. The results of experiments indicated that the difference in raw integrated density and manipulated pixels was a mere 0.0015% and 0.10% (17 pixels), respectively. These results show that the proposed technique has sufficient ability to distinguish the very slight manipulation. In a verification of the total power efficiencies of the CAM processor and three conventional mobile processors, we found that the value of Mbps/W, which is scaled to 45-nm CMOS technology for the morphological pattern spectrum and the AES cipher algorithm, drastically improved the power efficiency of the proposed image-manipulation detection with the CAM processor due to 16,382-way bit-serial and word-parallel operations. The CAM processor-based image manipulation detector achieves highly parallel processing with low power consumption; consequently, the proposed image-manipulation detection method with the CAM system is very effective for investigating crimes and for obtaining photographic evidence, especially in mobile products.