Today, the characterization of the technological age is done by the digital images spread. They are the most common form of conveying information whether through internet, newspapers, magazines, or scientific journals. They are used as a strong proof against various crimes and as evidence used for various purposes. The modification, capturing or creating of the image has become easier and available with the emergence of means of image editing and processing tools. One of the most important and popular type of image forgery is copy-move forgery in which an image part is copied and then pasted into the same image that has intention of hiding something important or showing a false scene. Because the important properties of the copied parts comes from the same image, such as brightness, noise, and texture which will be compatible with the entire image that makes more difficult for experts for the detection and distinguishing the alteration. Usually, the detecting copy move forgery conventional techniques suffer severely from the time-consuming problem. The evaluation of the improved method had been done using (150) images that was selected from two different datasets, “CoMoFoD” and “MICC-F2000”. Experimental results show that the improved method can accurately and quickly reveal the doubled regions of a tampered image. In addition, greatly reducing the processing time in comparison to the khan algorithm, and the accuracy is kept at the same level. Owing to the availability and technological advancement of the image editing sophisticated tools, there is increase in the loss of authentication in digital images. Thus, this led us to the proposal of different detection techniques that checks whether the digital images are forged or authentic. The specific type of forgery technique is copy move forgery in which widely used research topic is detection under digital image forensics. In this thesis an enhancement of copy move image forgery classification is done by implementing a hybrid features with classification algorithm like SIFT with SVM and EM algorithm and ORB with SVM and EM .The technique works by applying firstly the DCT on an image and then on an resultant image, SIFT is obtained after applying DCT. A supervised learning method is proposed for classifying a copy-move image forgery of TIFF, JPEG, and BMP. The process starts with reducing the color of the photos. Achieve the accuracy more than 90%.