Tampering identification using Empirical Frequency Response

With the widespread popularity of digital images and the presence of easy-to-use image editing software, content integrity can no longer be taken for granted, and there is a strong need for techniques that not only detect the presence of tampering but also identify its type. This paper focusses on tampering-type identification and introduces a new approach based on the empirical frequency response (EFR) to address this problem. We show that several types of tampering operations, both linear shift invariant (LSI) and non-LSI, can be characterized consistently and distinctly by their EFRs. We then extend the approach to estimate the EFR for scenarios where only the final image is available. Theoretical reasoning supported by experimental results verify the effectiveness of this method for identifying the type of a tampering operation.