Makeup-Free Biometrics Using Texture Analysis: Mitigating Imposture in Face Recognition | Chapter 9 | New Horizons of Science, Technology and Culture Vol. 5

A biometric authentication system is used for various applications, like the security of computers and mobile phones, airports, banks, military bases, biometric attendance, and tracking systems. Though biometric systems improve security, like any other system, they are vulnerable and prone to threats. Face recognition has quickly become one of the most common ways to authenticate using biometric data. Traditional face recognition systems mainly focus on extracting features and improving the accuracy of verification and identification. However, attention to security weak spots and possible attacks has only gained traction in recent years. These attacks include methods like obfuscation, spoofing, and morphing; for example, an intruder might impersonate a legitimate user to get around the system. Cosmetic changes can make recognition even harder by altering facial features like skin tone, eyebrow position, and overall complexion. These modifications can decrease the uniqueness of facial traits, often leading to false matches and weakening the security of the biometric system. To tackle this issue, adding a presentation attack detection (PAD) module to existing systems has been suggested. In this work, a CNN-based machine learning approach is used to classify presentation attacks through texture analysis. The proposed method removes makeup effects to restore the original look of the face, allowing the recognition system to identify individuals correctly and reduce the risk of attacks. The system's strength is measured using the False Accept Rate (FAR), which assesses its resistance to zero-effort attacks and serves as an important performance measure for biometric authentication systems. By training both pix2pix GAN and Cycle GAN and comparing their image quality, measure them. It is proven that Cycle GAN is more efficient than pix2pix GAN.  In the end, a comprehensive framework that combines various attack-prevention models can greatly enhance the strength and reliability of biometric authentication.

 

 

Author(s) Details

Logeswari Saranya R
Department of Information Technology, PSG College of Technology, Coimbatore, India.

 

Umamaheswari K
Department of Information Technology, PSG College of Technology, Coimbatore, India.

 

Please see the book here :- https://doi.org/10.9734/bpi/nhstc/v5/6510