A robust approach to detect video-based attacks to enhance security. (30th October 2019)

• Record Type: Journal Article
• Authors: Arora, Shefali; Bhatia, M.P.S.
• Is Part Of: International journal of innovative computing and applications. Volume 10:Number 3/4(2019)
• Page Start: 175
• Page End: 184
• Publication Date: 2019-10-30
• Subjects: video attacks -- replay attacks -- deep learning -- convolutional neural networks -- biometrics -- long short-term memory -- face spoofing -- real time detection -- security -- innovative computing; Evolutionary computation -- Periodicals; Neural networks (Computer science) -- Periodicals; Genetic programming (Computer science) -- Periodicals; Biologically-inspired computing -- Periodicals; Swarm intelligence -- Periodicals; Quantum computers -- Periodicals; 006.3
• Languages: English
• ISSNs: 1751-648X

Abstract

Face authentication has become widespread on smart devices and in various applications these days. Real time detection of human faces in video surveillance systems is challenging due to variations in expression and background conditions. Thus, precise detection of spoofed faces is important to make such security systems robust against potential attacks. Several deep learning-based techniques involving the use of convolutional neural networks have proven to be excellent in detection of spoofed faces. In this paper, we have proposed the combined use of spatial and temporal information from facial images using CNN and long short-term memory networks. We have tested the approach on Idiap Replay Attack benchmark and compared the results with the application of pre-trained models like InceptionV3, VGGNet and ResNet models to detect replay attacks during video surveillance. Our approach proves to be robust and more efficient for detection of security breaches in real time situations.