FACIAL IDENTIFICATION FOR DIGITAL FORENSIC

Abstract

Forensic facial recognition has become an essential requirement in criminal investigations as a result of the emergence of electronic devices, such as mobile phones and computers, and the huge volume of existing content. Forensic facial recognition goes beyond facial recognition in that it deals with facial images under unconstrained and non-ideal conditions, such as low image resolution, varying facial orientation, poor illumination, a wide range of facial expressions, and the presence of accessories. In addition, digital forensic challenges do not only concern identifying an individual but also include understanding the context, acknowledging the relationships between individuals, tracking, and numbers of advanced questions that help reduce the cognitive load placed on the investigator. This thesis proposes a multi-algorithmic fusion approach by using multiple commercial facial recognition systems to overcome particular weaknesses in singular approaches to obtain improved facial identification accuracy. The advantage of focusing on commercial systems is that they release the forensic team from developing and managing their own solutions and, subsequently, also benefit from state-of-the-art updates in underlying recognition performance. A set of experiments was conducted to evaluate these commercial facial recognition systems (Neurotechnology, Microsoft, and Amazon Rekognition) to determine their individual performance using facial images with varied conditions and to determine the benefits of fusion. Two challenging facial datasets were identified for the evaluation; they represent a challenging yet realistic set of digital forensics scenarios collected from publicly available photographs. The experimental results have proven that using the developed fusion approach achieves a better facial vi identification rate as the best evaluated commercial system has achieved an accuracy of 67.23% while the multi-algorithmic fusion system has achieved an accuracy of 71.6%. Building on these results, a novel architecture is proposed to support the forensic investigation concerning the automatic facial recognition called Facial-Forensic Analysis System (F-FAS). The F-FAS is an efficient design that analyses the content of photo evidence to identify a criminal individual. Further, the F-FAS architecture provides a wide range of capabilities that will allow investigators to perform in-depth analysis that can lead to a case solution. Also, it allows investigators to find answers about different questions, such as individual identification, and identify associations between artefacts (facial social network) and presents them in a usable and visual form (geolocation) to draw a wider picture of a crime. This tool has also been designed based on a case management concept that helps to manage the overall system and provide robust authentication, authorisation, and chain of custody. Several experts in the forensic area evaluated the contributions of theses and a novel approach idea and it was unanimously agreed that the selected research problem was one of great validity. In addition, all experts have demonstrated support for experiments’ results and they were impressed by the suggested F-FAS based on the context of its functions.