Video Inter-frame Forgery Detection Approach for Surveillance and Mobile Recorded Videos

Abstrak

We are living in an age where use of multimedia technologies like digital recorders and mobile phones is increasing rapidly. On the other hand, digital content manipulating softwares are also increasing making it easy for an individual to doctor the recorded content with trivial consumption of time and wealth. Digital multimedia forensics is gaining utmost importance to restrict unethical use of such easily available tampering techniques. These days, it is common for people to record videos using their smart phones. We have also witnessed a sudden growth in the use of surveillance cameras, which we see inhabiting almost every public location. Videos recorded using these devices usually contains crucial evidence of some event occurence and thereby most susceptible to inter-frame forgery which can be easily performed by insertion/removal/replication of frame(s). The proposed forensic technique enabled detection of inter-frame forgery in H.264 and MPEG-2 encoded videos especially mobile recorded and surveillance videos. This novel method introduced objectivity for automatic detection and localization of tampering by utilizing prediction residual gradient and optical flow gradient. Experimental results showed that this technique can detect tampering with 90% true positive rate, regardless of the video codec and recording device utilized and number of frames tampered.