The College of Education for Pure Sciences, Department of Computer Science, discussed a master’s thesis on detecting facial impersonation using the CNN neural network.
The thesis submitted by researcher (Hala Shaker Mahmoud) included presenting a method to detect attacks on face recognition systems used in security and privacy technologies. This method uses a variety of methods and data provided by some cameras and sensors to detect facial impersonation using convolutional neural networks (CNN) in the context of deep learning. To evaluate the effectiveness of the proposed approach in real-world scenarios, the Presentation Multi-Channel Attack (WMCA) and 3D Mask Attack (3DMAD) datasets were used. The presented method exploits multi-modal data, including color images, 3D data, infrared images, and thermal, to enhance system performance and explore different techniques to combine results from each method. This study explores various techniques to combine results from each channel in two merger scenarios, pre-merger and post-merger. The former combines data from the four channels and then outputs a single face classification result, while the latter combines the results of each method. Various fusion techniques are used, including majority voting, weighted voting, mean clustering, and stacking classifier. This proposed system has shown commendable performance when compared with state-of-the-art methodologies.