Time-Aware Face Anti-Spoofing with Rotation Invariant Local Binary Patterns and Deep Learning

Moritz Finke,Alexandra Dmitrienko

2024-08-27

Abstract:Facial recognition systems have become an integral part of the modern world. These methods accomplish the task of human identification in an automatic, fast, and non-interfering way. Past research has uncovered high vulnerability to simple imitation attacks that could lead to erroneous identification and subsequent authentication of attackers. Similar to face recognition, imitation attacks can also be detected with Machine Learning. Attack detection systems use a variety of facial features and advanced machine learning models for uncovering the presence of attacks. In this work, we assess existing work on liveness detection and propose a novel approach that promises high classification accuracy by combining previously unused features with time-aware deep learning strategies.

Computer Vision and Pattern Recognition

What problem does this paper attempt to address?

### Problems the Paper Aims to Solve

The paper primarily aims to address the issue of Presentation Attack Detection (PAD) in facial recognition systems, specifically:

1. **Importance of Attack Detection**:
- Facial recognition systems have become an indispensable part of the modern world, but these systems are vulnerable to simple spoofing attacks (such as printed photos or video replays), leading to incorrect identity verification and subsequent security threats.

2. **Limitations of Existing Methods**:
- Although existing attack detection methods can already utilize various facial features and advanced machine learning models to identify attacks, they still face challenges in practical applications due to resource constraints, user experience requirements, and data processing speed.

3. **Proposed New Method**:
- This paper proposes a new liveness detection system—Color and Texture LSTM (C&T-LSTM), which aims to achieve high-precision classification while maintaining fast authentication by combining previously unused features with a time-aware deep learning strategy, and is highly compatible with end devices.

In summary, the main goal of the paper is to enhance the security of facial recognition systems against various forms of spoofing attacks while ensuring user convenience and system compatibility.