A Novel Framework to Detect Image Manipulation Using Pixel and Lighting

The fast-paced technological developments have rendered the role of digital images very common in the present scenario. Not only the technology is used to savour and store our memories of events, gatherings, news and other memorable moments, but also used digital images and digital media as a proof of crime, accident, and other such events for justice and legal actions. This comes from the most fundamental concept of "we trust what we see". However, this fact is also abused in terms of photo forgeries. With digital computers and image-editing software, it's simpler to change digital images. This includes fabricated digital photos that are presented as evidence in court or are used to recreate a crime scene. Authenticating such evidence is crucial in such situations in order to fairly resolve the case. For example, according to Wall Street Journal, approximately 10% photographs published in the USA are digitally reformed. Even original scientific reports are also been altered to modify the original data.

Sometimes forgeries may not be detected by using conventional image verification methods, such as, naked eyes, histogram, comparison of pixel values etc.. Hence, there is a requirement of a more reliable tool to identify a forgery more accurately. These tools are usually based on physical and mechanical properties of digital images and sensors, respectively, to detect any abnormality in an image. Furthermore, image manipulation tools evolve with latest technical developments, and hence, these forgery detection methods require to be more robust and more reliable.

Although digital image forgery detection and security of the information in a digital image (steganography) may look like the same process, however the two are different in nature. While the first one deals with identification of image manipulation without knowing any changes in base images, whereas the second one requires hiding information in base images to identify any changes happening to the images later on.