528 IJSR - INTERNATIONAL JOURNAL OF SCIENTIFIC RESEARCH Volume : 4 | Issue : 10 | October 2015 • ISSN No 2277 - 8179 Research Paper Computer Science K.Alice Asst Prof ,GKM College of Engineering and, Technology, Anna University N.Ramaraj Professor, hanagavelu Engineering College, Anna University Combining Hashing Techniques in Image Authentication system:A Survey KEYWORDS : Content based Image au- thentication ,DWT and SVD, AMAC ,Fixed Point heory, HMM and SVM ABSTRACT In content based image authentication system the image hash is generated by extracting image features that holds the content of the image using transform coeicients, Edges, Statistical methods, Local and Global descriptors, Histo- gram etc. Many authentication systems employ any one of the feature extraction methods to generate an image hash. Recently most authen- tication system uses a combination of two or more feature extraction methods, to extract the content of image and then to generate an image hash. his is done to fully exploit the advantage of two or more methods over just a single method. In this paper a survey on most recent work in the area of content based image authentication that uses two or more feature extraction techniques and a performance analysis on each of the system based on the requirement of an authentication system is given. 1. Introduction he widespread use of multimedia data in internet increases the need for image authentication. Image authentication veriies the originality of the transmitted image at the receiver end. Image authentication is highly essential in the area of military target images, court evidence images, quality control images, digital notaries’ document, medical images, etc where false judgment leads to tragic results. Multimedia data authentication is difer- ent from other data authentication because multimedia data is always preprocessed (Quantized, Compression, Scaling or Crop- ping) before usage and transmission [2]. Image authentication is broadly classiied into two categories: Strict (Hard) authentication and Selective (Soft or Content based) authentication [1]. Strict authentication cannot tolerate even a single pixel change in image. his authentication is im- plemented by using conventional cryptographic algorithm (SHA- 1, MD5 etc) and fragile watermarking techniques. he basic idea behind fragile watermarking technique is use to generate a wa- termark and it is inserted in the image in such a way that any modiication in the image is relected in the inserted watermark. Simply verifying the presence of the inserted watermark allows the image authenticity veriication. he image is considered au- thentic if and only if its entire pixel remains unchanged. Selective authentication can be classiied as digital signature [7] and semi-fragile watermarking [8]. Selective authentication method authenticates the image as long as the content of the image is preserved during the modiication. Some of the content preserving modiication includes image ile format, geometric transformations, (scaling, translation, rotation etc) transmis- sion errors, transmission noise, compression, quantization etc. his method fails in authentication for those modiications that change the content of image. Some of the manipulation that change the image content may be deletion of objects from the image, addition of objects into the image, position change of the object in the image , change in image background, change in lu- minous etc. Content based image authentication concentrates on authenticat- ing the content of the image rather than image as a whole. Most approaches extract the features of the image that holds the con- tent of the image and provide security and authenticity. Features can be extracted based on transform coeicients’ (DCT, DWT, SVD, etc), edges, histogram, local and global descriptors etc. hese features are used to create image hash for authentication purpose. The basic requirement of a good authentication system can be given as follows [1]: Robustness- The system should tolerate content preserving transformations. Security- The system should be able to protect the data from malicious attack. Sensitivity- The system should detect any content modiications or manipulations. Localization - The system should be able to locate the area of tampering. Recovery- The system should be able to reconstruct the tampered regions. Complexity - The system must be neither complex nor slow. Portability- The system must be able to carry the digital signature within the image. he paper is organized as follows: Section II describes the meth- ods of few recent works in the area of image authentication. Sec- tion III provides a performance analysis of all the methods de- scribed in section II based on the requirement of authentication system. Section IV concludes the paper. 2. Few recent work in the area of content based image au- thentication Two phase image authentication Two phase image authentication method [2], uses a combination of strict and selective authentication using two most generic ap- proximate message authenticate codes (AMAC). Image feature that contains the content of image are extracted and are protect- ed by hard authentication using AMAC which enables forgery de- tection and the less fragile content holder parts are protected by soft authentication to provide robustness to the mechanism Discrete cosine transformation is applied to equal sized blocks of the image and the low frequency elements of the image is subjected to irst phase to generate tag1 using DCT-RS( Reed Solomon) based scheme. If the irst phase succeeds, some high frequency of the im- age are left for the second phase to generate tag 2 using probabil- istic target collision and resistant hash function. Tag 1 and tag 2 is concatenated to generate digital signature for the image. MAC is sensitive to noise but AMAC is noise tolerant. he use of AMAC increases robustness. AMAC enables error detection and partial error localization. he proposed method is unable to classify and identify the type of attack or to localize the afected part of the image. Watermarking for image based on DWT & SVD he proposed method [3] generates a digital watermark by combin-