In this paper, we aim to address the image recapturing detection problem with the convolutional and recurrent neural networks. With the advances of image display and acquisition techniques, the recaptured images are of satisfactory quality nowadays. This has been creating an ever stronger demand for sophisticated image recapturing detection algorithms which can efficiently prevent the unauthorized image distributing and forging. In this paper, we propose a hierarchical feature learning strategy by leveraging the intra-block information and inter-block dependency for image recapturing detection. In particular, the image blocks are first employed as the input of the convolutional neural network (CNN) and subsequently recurrent neural network (RNN) is further adopted to extract block dependencies. The CNN and RNN serve as effective tools to extract discriminative and meaningful features regarding both intra- and inter-block information. As such, the inherent properties within local blocks and the correlations between non-local neighbouring blocks are all exploited to identify the recaptured images. Experimental results on three databases show that significantly better performance can be achieved with our proposed framework compared to traditional handcrafted and deep learning based approaches.
Video forgery detection is becoming an important issue in recent years, because modern editing software provide powerful and easy-to-use tools to manipulate videos. In this paper we propose to perform detection by means of deep learning, with an architecture based on autoencoders and recurrent neural networks. A training phase on a few pristine frames allows the autoencoder to learn an intrinsic model of the source. Then, forged material is singled out as anomalous, as it does not fit the learned model, and is encoded with a large reconstruction error. Recursive networks, implemented with the long short-term memory model, are used to exploit temporal dependencies. Preliminary results on forged videos show the potential of this approach.