Research And Implementation Of Image Steganalysis System Based On Spark

With the advent of big data,there has been an explosive growth of image files.In recent years,images was used to conceal malicious software by using steganography and spread widely across the network,which causing tens of millions of users to suffer from cyber attacks such as ransomware and privacy leakage.Even more dangerous,steganography tools and and the number of cyber attacks using steganography have increased rapidly.However,due to the large scale of images in the network,the computational performance of the traditional steganalysis technology is limited.This makes it difficult to transform steganalysis into real scenarios.In view of the above problems,this thesis proposes an image steganalysis system based on Spark,aiming to solve the problem of large-scale image steganalysis and provide users with practical and easy-to-use services.The main work includes:(1)Aiming at the problem of limited performance of traditional steganalysis algorithms or algorithms based on MapReduce framework in the face of large-scale images,this thesis redesigns steganalysis algorithm based on Spark calculation and data storage mode after studying steganalysis algorithm and analyzing parallelizable parts.The steganalysis algorithm based on Spark performs image steganalysis tasks in parallel,and experiments show that the parallel algorithm is more efficient than the traditional steganalysis algorithm.(2)Aiming at the problem of detecting accuracy of steganalysis algorithm in the face of complex image sources,this thesis researchs the key parameters determining the performance of the image steganalysis algorithm.By using multi-set SRM parameters and comparing the two approaches by specifying ensemble classifier parameters and automating the search for the best parameters,this thesis carried out experiments to evaluate steganalysis algorithm performance,so as to determine the appropriate algorithm parameters in the Spark cluster constructed in this thesis.Experiments show that the detection accuracy of the improved algorithm is higher than that of the existing steganalysis algorithm.(3)Aiming at the problem of there is a lack of steganalysis tools for practical application scenarios at present,this thesis researchs the latest Web development technology and designs and implements the image steganalysis system based on Spark with universality,high efficiency and ease of use.The system mainly includes user management module,image set management module,parallel steganalysis based on Spark module and report management module.(4)This thesis tests the interface,function and performance of the system.The average error rate and acceleration ratio are used to measure the detection accuracy and parallelization performance of the improved algorithm.Experiments show that the accuracy and efficiency of the algorithm can meet the requirements of steganalysis of large-scale images.