Machine Learning Approach to Detect Tampering in H.264 Video
Remya R. S1, Anupama Pradeep2
1Asst. Prof. Remya R. S, Department of Computer Science, College of Engineering, Karunagapally, Kollam (Kerala), India.
2Anupama Pradeep, PG Scholar, Department of Computer Science, College of Engineering, Karunagpally, Kollam (Kerala), India.
Manuscript received on 14 June 2015 | Revised Manuscript received on 23 June 2015 | Manuscript Published on 30 June 2015 | PP: 81-85 | Volume-5 Issue-1, June 2015 | Retrieval Number: A2126065115/15©BEIESP
Open Access | Editorial and Publishing Policies | Cite | Mendeley | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: Now a days there are plenty of software’s available to access and edit digital videos. Therefore video tampering detection is crucial for legal, medical and surveillance applications. Digital videos are considered as more reliable source of evidence than still images. The abundance of compressed video forms a potential thread of evidence in court rooms. In case of artifacts and possibility of fraud videos court usually calls forensic investigators for examining the problem of authenticating multimedia content. An automated objective assessment of digital video helps to increase the accuracy of videos. Existing schemes are based on MPEG codec. This paper proposes a novel technique to detect tampering in H.264 videos by using neural network. This paper identifies video tampering by using a feature called sequence of average residual of P-frames (SARP). Then time and frequency domain features of sequence of average residual of P-frames are calculated. The detection system is trained with these features. Then the detection system is applied to the video sequence under examination. This method identifies video tampering by differences in time domain and frequency domain features of tampered video from original video. By using machine learning approach, it classifies type of tampering such as insertion, deletion and copy-move. PNN is used for training. The proposed method is applicable for different codec.
Keywords: Video Tampering Detection, SARP, Time Domain Feature, Frequency Domain Feature, Training.
Scope of the Article: Machine Learning