A faster and intelligent steganography detection using Graphics Processing Unit in cloud
Mayank Tiwary, Rojalina Priyadarshini, Rachita Misra · 2014
Now a day's security in big data is a major concern and there is always a threat to the integrity of data. Contents of the data are often challenged by several external attacks. We cannot deny the fact of presence of unnecessary data in form of secret messages. Through steganographic techniques, a text messages can be hidden inside the images, video files, mp3's, etc. So, there is always a need to store disinfected data in cloud database. But researchers find it very difficult to differentiate between infected and disinfected data, when the size of data is huge; it takes lot of time to process the infected data and remove the noises in conventional uniprocessing system is used. This paper introduces faster and artificially intelligent steganalysis techniques based on parallel algorithms. The technique involves image estimation based on order statistic filters, feature generation from the estimated image and finally classifying the generated images into stego and non-stego type. An artificial neural network is used as a machine classifier on the generated images to detect either stego or clean images. The whole process in a single processor environment is computationally intensive. Therefore the technique offloads the calculations to the high performance Graphics Processing Units for faster operations. This piece of work uses the Compute Unified Device Architecture (CUDA) for the integration process.