Tampering Detection of Audio-Visual Content Using Encrypted Watermarks

Abstract

In this paper, we present a framework for detecting tampered information in digital videos. Using the proposed technique is possible to detect several types of tampering with a pixel granularity. The framework uses a combination of temporal and spatial watermarks that do not decrease the perceived quality of the host videos. We use a modified version of Quantization Index Modulation (QIM) algorithm to store the watermarks. Since QIM is a fragile watermarking scheme, it is possible to detect local, global, and temporal tampers and also estimate the attack type. The framework is fast, robust, and accurate.