An accurate multi-modal biometric identification system for person identification via fusion of face and finger print

Internet of things (IoT) have entirely revolutionized the industry. However, the cyber-security of IoT enabled cyber-physical systems is still one of the main challenges. The success of cyber-physical system is highly reliant on its capability to withstand cyberattacks. Biometric identification is the key factor responsible for the provision of secure cyber-physical system. The conventional unimodal biometric systems do not have the potential to provide the required level of security for cyber-physical system. The unimodal biometric systems are affected by a variety of issues like noisy sensor data, non-universality, susceptibility to forgery and lack of invariant representation. To overcome these issues and to provide higher-security enabled cyber-physical systems, the combination of different biometric modalities is required. To ensure a secure cyber-physical system, a novel multi-modal biometric system based on face and finger print is proposed in this work. Finger print matching is performed using alignment-based elastic algorithm. For the improved facial feature extraction, extended local binary patterns (ELBP) are used. For the effective dimensionality reduction of extracted ELBP feature space, local non-negative matrix factorization is used. Score level fusion is performed for the fusion. Experimental evaluation is done on FVC 2000 DB1, FVC 2000 DB2, ORL (AT&T) and YALE databases. The proposed method achieved a high recognition accuracy of 99.59%.