An Enhanced Eigenfaces-based Biometric Forensic Model
The recent explosive development of the Internet allowed unwelcomed visitors to gain access to private information and various critical - mission resources such as financial institutions, hospitals, airports ... etc. Internet security has become a hot topic and relies on advanced technology. Now, more than ever, there is an increasing need for stronger identification mechanisms such as biometrics, which are in the process of replacing traditional identification solutions. Also, critical - mission systems and applications require mechanisms to detect when legitimate users try to misuse their privileges. Biometrics enables cybercrime forensics specialists to gather evidence whenever needed. This paper aims to introduce a biometric forensic model using facial identification approach. This model is based on the Eigenfaces approach for recognition proposed by Turk and Pentland . Here, an unknown input image is compared with a set of images stored in a database to identify the best match. A freely accessible faces database has been used to develop our model which is based on a mathematical approach, called Principle Component Analysis (PCA). The paper addresses the issue of extracting global features of the images which are stored separately in the database. The features of a test image were compared with a set of images whose features were stored. The distance of the two images was calculated and when was minimum and below a certain threshold, the two images were considered to be the same and belong to a particular person. The calculated distance could be used and / or adjusted by a forensic specialist for deciding whether or not a suspicious user is actually the person who claims to be. The performance of the proposed face identification model was evaluated using standard methods. Distance values were used to express the similarity between any input image and other stored images. The model’s performance was evaluated using FAR (False Acceptance Rate), FRR (False Rejection Rate) and EER (Equal Error Rate). In FAR, each user’s image was compared with all images present in the database excluding the user’s own image. In FRR, each user’s image was compared with his own stored in the database. The major findings of the experiments showed promising and interesting results in terms of the model’s performance and similarity measures.