Advancing State-of-the-art of Biometrics Presentation Attacks and their Detection

Due to a large surge of illegal access, authentication of access has gained significant attention. Biometrics recognition is one of the most secure metrics towards that goal. While the metric achieved significant performance due to advancement in machine learning and artificial intelligence algorithms, the system is still highly vulnerable against presentation attacks also known as spoofing.

More precisely, a presentation attack refers to the presentation of an artifact or human characteristic to the biometric capture subsystem in a manner that can interfere with the envisioned policy of the biometric system. As the deployment of biometric systems increases, the significance of effective security measures against presentation attacks becomes vital. Therefore, presentation attack detection has received much attention from the computer vision and pattern recognition communities. Due to the recent advances in deep learning and big data analytics, significant research is being performed in the related fields.

In recent years, several research works have been published advancing the knowledge towards the development and detection of the presentation attack instrument. However, the security field is a game of cat and mouse, where one tries to advance the attack, another tries to break the security. Beyond general-purpose classification systems, other examples testify to the existence of adversarial attacks. Security algorithms, such as presentation attack detection algorithms, are found vulnerable and require further attention. Similarly, the new attacks come into the picture such as wax figure faces which makes the earlier deployed detection algorithms less effective.

Therefore, we welcome original research papers making substantial theoretical and practical contributions on robust presentation attack detection in combination with computer vision and machine learning topics, including, but not limited to:
● Novel methodologies on presentation attack detection in visual media
● Studies on novel attacks to biometric systems and solutions
● Zero-shot learning for presentation attack detection
● Deep learning methods for biometric authentication systems using visual media
● Novel datasets and evaluation protocols on spoofing prevention on visual and multimodal biometric systems
● Generative models (e.g. GAN) for presentation attacks
● Novel methods for detecting and preventing presentation attacks
● Adversarial vulnerability of the presentation attack detection algorithms

Keywords: Presentation Attacks, Biometrics, Spoofing, Robustness, Demographic Bias, Adversarial Attacks

Submission Deadlines