An end-to-end deep source recording device identification system for Web media forensics

Abstract

Purpose Most source recording device identification models for Web media forensics are based on a single feature to complete the identification task and often have the disadvantages of long time and poor accuracy. The purpose of this paper is to propose a new method for end-to-end network source identification of multi-feature fusion devices. Design/methodology/approach This paper proposes an efficient multi-feature fusion source recording device identification method based on end-to-end and attention mechanism, so as to achieve efficient and convenient identification of recording devices of Web media forensics. Findings The authors conducted sufficient experiments to prove the effectiveness of the models that they have proposed. The experiments show that the end-to-end system is improved by 7.1% compared to the baseline i-vector system, compared to the authors’ previous system, the accuracy is improved by 0.4%, and the training time is reduced by 50%. Research limitations/implications With the development of Web media forensics and internet technology, the use of Web media as evidence is increasing. Among them, it is particularly important to study the authenticity and accuracy of Web media audio. Originality/value This paper aims to promote the development of source recording device identification and provide effective technology for Web media forensics and judicial record evidence that need to apply device source identification technology.

Spatial and temporal learning representation for end-to-end recording device identification

“…This work is an extension of our previous work [14]. In general, this work has the following contributions:…”

“…However, the fitting of a shallow network does not fully reflect the effect of deep learning. Further, Chunyan Zeng [14] et al used a multi-feature parallel convolution network, combined with the attention mechanism for device source identification to achieve an improved effect. With further research on device source identification, the feature dimension of device source identification and the number of devices for device source identification tasks will further increase.…”

“…In the feature fusion module, we use the attention mechanism to fuse the device source spatial information and temporal information to obtain more representative features. Based on our previous work [14], we used full connection with a decreasing number of nodes in the back-end classification module in this paper. In the deep-and-shallow loss module, we designed a deep-and-shallow loss for the end-to-end framework, and the specific information is described in Section 3.2.…”

“…The DNN network with decreasing neurons can effectively reduce parameter redundancy and help deploy deep learning applications. The parameter designed according to our previous work [14].…”

“…LSTM (Long Short-Term Memory) can reasonably handle the temporal correlation in sequential data [13], which has certain advantages in the sequence model with long-term memory. In addition, the deep learning methods directly make predictions based on the input data, which are conducive to the design of an end-to-end framework for device source identification tasks [14]. Deep learning has many advantages, but device source identification based on deep learning still has many challenges.…”