Article Preview
Top1. Introduction
Undeterred by the fact that biometric systems offer reliable techniques for personal identification, their usage could be hampered by the deficiency of a proper protection scheme that guarantees the security and privacy of the biometric traits. When biometric images or templates are transmitted through insecure channels or stored as raw data, they run risks of being stolen or modified. Hence, it is imperative that robust and reliable means of biometric protection are implemented. Ratha et al. (2001) describe eight types of attacks that are possible in a biometric system, such as database template tampering, template modification, the matcher override of the final decision and attack on the channel between the feature extractor and the matcher, or attack on the channel between the database and matcher. Moreover, due to the extensive usage of biometrics technology in many applications, it is very likely that biometric data are being transmitted over non-secure channels. Hence, for a biometric system to work properly, the system must guarantee that the biometric data came from a legitimate person at the time of enrolment. Several means are employed to protect biometric data such as only encryption, or watermarking and encryption. Encryption can be used as one potential mechanism for protecting the biometric features (Martin et al., 2009); however, encryption may limit the capacity of large scale biometric systems because it can be computationally expensive. In addition, encryption cannot provide complete protection as the templates must be decrypted before matching (Jain et al., 2001). emphasized this by suggesting that if only cryptographic techniques are used for the protection of biometric data, security of such data is not fully maintained because this data has to be decrypted somewhere. Therefore, the use of watermarking technology has emerged. Since watermarking involves hiding information within the host data, it can provide security even after decryption. On the other hand, Visual Cryptography (VC) can be utilized for biometric template protection. This template is usually stored as raw data in databases or transmitted over unsecured channels matching the reconstructed synthetic images against the original ones. Applications for face recognition can be found in the areas of entertainment, smart cards, information security, law enforcement, medicine, and security (Sundaram, Pradeep, Vignesh et al., 2015; Ratha et al., 2001). Diverse techniques and systems have been created for face detection in areas that use cameras in the visible spectrum. Machine recognition of human faces has experienced great strides but remain challenged by intricate issues related to light variability (Martin et al., 2009) and other factors like difficulty in detecting facial disguises. The use of thermal mid-wave infrared (MWIR) portion of the electromagnetic (EM) spectrum solves the problem of light variability. Also, any foreign object on a human face such as a fake nose could be detected, as foreign objects have a different temperature range than that of human skin. Due to these benefits, a lot of effort has been aimed at developing human face recognition systems in the MWIR spectrum. However, since cameras in the MWIR portion of EM spectrum are available at a much higher cost than their visible band counterparts, much of the research done in human face recognition in the MWIR spectrum is still in its infancy. In recent years, researchers have realized the potential of thermal MWIR imagery for human identification using the vein structure of hands (Jain et al., 2005; Daugman et al., 2004) finger vein patterns (Venugopalan et al., 2011; Galbally et al., 2013) and vein structure of the human face (Park et al., 2007). Thermal images have been used to identify the affective state of humans (Hassanien et al., 2009). Moreover, the fusion of visual and thermal face images has been used in the face recognition field. Given the complex nature of human vasculature, this approach to face recognition using MWIR imaging is checked against another existing database to prove the reliability of the algorithms designed for feature extraction, template generation, and authentication through similarity measures. The measure of performance is based on human vasculature which are unique for an individual hence success rate is high because only matched face with the existing database is to be considered further.