OPTIMAL FACE SPOOF DETECTION BASED ON IMPROVED WHALE OPTIMIZATION ALGORITHM AND DUAL STAGE CONVOLUTIONAL NEURAL NETWORK HYBRID

Abstract

Face detection systems are now being utilized in various application systems, for instance, periphery crossing points, banks, and security information access etc. Advanced facial verification systems can focus on the vulnerability of facial biometrics to process swindlers and use photos, videos or 3D nodes of verified customer faces to create access to the workplace or system and important secured data. Facial correction or dynamic differential correction algorithms (e.g., face activation or deception selection) could be utilized in solving such a crisis of deception. Adversely, the current issue is due to the difficulty in the detection of bias and computation anomaly intrusions. The paper proposes an optimal face spoof detection (OFSD) methodology along with artificial intelligence framework for the spoofing detection procedure. Initially the methodology of proposed OFSD design begins with a dual stage convolutional neural network (DS-CNN) managing different comparisons of space information for face spoof recognizable proof. Secondly, an improved whale optimization algorithm (IWOA) streamlining estimation for feature classification to select best perfect features from various level is utilized. Subsequently an optimal rule based fusion (ORBF) procedure for sufficient interweaving of the features from different sources is structured. Exploratory analysis of three publicly available databases, REPLAY-ATTACK, CASIA-FA and OULU-NPU, reveals interesting results that differ from previous work.