Face anti-spoofing detection based on DWT-LBP-DCT features

doi:10.1016/j.image.2020.115990

Signal Processing: Image Communication

Volume 89, November 2020, 115990

https://doi.org/10.1016/j.image.2020.115990 Get rights and content

Highlights

•
Propose a new face anti-spoofing detection approach based on DWT-LBP-DCT Features with a SVM classifier.
•
Investigate functions of DWT and DCT in the approach in both experimental and analytical ways, respectively.
•
The proposed scheme achieves better detection accuracy and lower detection time.

Abstract

In this paper, we propose a face anti-spoofing strategy by using DWT (Discrete Wavelet Transform), LBP (Local Binary Pattern) and DCT (Discrete Cosine Transform) with a SVM classifier to evaluate whether a video is valid. Firstly, the DWT features are produced by decomposing some selected frames into different frequency components at the $8 * 8$ multi-resolution blocks. Secondly, the DWT-LBP features are generated to represent spatial information of the blocks by connecting LBP histograms of the DWT blocks in each frame horizontally. Then, the DWT-LBP-DCT features with the temporal information of a video file are achieved by performing DCT operation on the DWT-LBP features of those selected frames vertically. As a result, these exploited DWT-LBP-DCT features have the capacity to represent the frequency-spatial–temporal information of a video. Finally, the SVM classifier with RBF kernel is trained for face anti-spoofing. Compared with previous excellent works, experimental results on two benchmark databases (REPLAY-ATTACK and CASIA-FASD) have demonstrated the proposed approach has better detection performance.

Introduction

In recent years, the prosperous development and widespread application of intelligent technologies have placed more stringent requirements on the safety of biometric authentication systems. Different biometric authentication systems used different physical or behavioral characteristics to identify persons [1]. Alternatives have been explored over the years, including iris [2], voice [3], fingerprint [4], face information [5], and handwriting signatures [6]. Among them, face information becomes one of the best choices for system authentication from the perspective of low recognition cost and convenience. At present, the maturity of identification technology has enabled the biometric authentication system to be applied in many occasions, ranging from access control systems for various secret occasions to landing systems for mobile terminals, and even to the face recognition technology in mobile payment systems [7]. Face biometrics have achieved remarkable performance over the past decades, but unexpected spoofing poses a threat to information security [8].

Compared with other biological features (such as iris and fingerprint), pictures and videos of legal users can be easily obtained through social networks. The low-cost characteristics make fraud behaviors for face recognition especially easy to be implemented. Illegal users can imitate the effective biometrics of a legal user through the camera. Therefore, the robust face anti-spoofing algorithm is particularly important in face recognition technology. Only when the face is judged to be living and belongs to a legal user, recognition result is true and effective; otherwise, this is determined as an illegal attack on the authentication system [7]. Under this double guarantees, the security and reliability of the system can be improved. As a result, improving the anti-spoofing ability of the face recognition system has become an urgent problem in face authentication [7]. The effective face algorithm is not only important for system security, but also has research value and broad prospects.

In general, face attacks can be divided into three categories: attacks by photos, attacks by videos, attacks by masks. Photo attacks mean an illegal user prints the photo as a paper version or displays it on an electronic device, and presents the photo to the camera of verification system. A video attack is that an illegal user replays a video belonging to a legitimate user and attacks the face recognition system with dynamic information. Mask attacks refer to an illegal user wearing the 3D mask of the original user, imitating the stereo effect of the face. In attacks by photos and videos, there is a possibility of artificial modification, that is, the background in the picture may be the original background, but the foreground image (face area) is generated by software simulation or manually replaced. The APP software ZAO which replaces face in videos with another face provided by users, demonstrates that face anti-spoofing technology has gradually matured and there are also various potential unknown forms of attack in the future. Considering all of these, we put our scope on photos and video attacks in this paper, and focus on a generalizable method.

In terms of the attacks, traditional hand-engineered features have been proposed to describe liveness. Texture characterization was often envisaged by measuring the fluctuations (with respect to space) of image amplitude regularity [9], and made great results like Local Binary Pattern (LBP) and Dynamic Mode Decomposition (DMD). In the meantime, with the improvement of computer computing performance and the development of deep learning techniques, the performance of deep learning methods are gradually catching up with the traditional methods in the detection.

Compared to previous work, the contribution of this paper can be summarized as:

•
Discrete Wavelet Transform (DWT) is utilized to present multi-resolution and location information by dividing image into coarse-scales and fine-scales.
•
Local Binary Patterns (LBP) features is performed on the DWT features to identify the texture and amplify the difference between true/fake frame blocks.
•
Discrete Cosine Transform (DCT) is applied on the DWT-LBP features to obtain main energy components of a video.
•
A SVM classifier is trained with the DWT-LBP-DCT features and tested by using intra-test and combined-test strategies in two benchmark databases. Experiment results have shown that the proposed method can provide higher detection accuracy.

The rest of the paper is organized as follows. We briefly review the state-of-art literatures of face anti-spoofing detection in Section 2 and explain the proposed approach in detail in Section 3. Then, two databases and evaluation protocols are introduced in Section 4. This is followed by a report of experimental results in Section 5. Finally, conclusions and future work are given in Section 6.

Section snippets

Related work

Feature extraction and classifier designing are two key steps for reliable detection in images. Reviewing the relevant research, the methods mainly used for extracting features for face attacks can be divided into five parts: static texture based, motion based, frequency based, color based and deep learning based. And the majority of traditional face detection methods combines manual features and a simple machine learning method to construct an anti-spoofing system.

Static texture based:

Proposed method

In the paper, we present a DWT-LBP-DCT features based approach which has combined the characteristics of DWT, LBP and DCT operations to form a high-level feature for face anti-spoofing attacks. The flowchart is shown in Fig. 1. The pipeline is to extract frames at first. After face recognition and area enlargement operation, we have performed a DWT operation on each frame. All input and output images for DWT are RGB images, which are converted to gray versions for LBP operations. After a DWT

Face anti-spoofing databases

In order to evaluate the performance of the proposed face anti-spoofing algorithm, we adopt two benchmarking databases, which are CASIA-FASD database and REPLAY-ATTACK database.

Experimental results

In this section, we tested the performance of the proposed DWT-LBP-DCT method by using two standard benchmark databases. Also, we analyzed functions of the DCT and DWT operations in the proposed method by applying the DWT-LBP-DCT, the DWT-LBP features and the LBP-DCT features for face-spoofing, respectively. The LBP-DCT features have been applied in our previous work [7].

Conclusion and future work

The paper is aimed to address the issue of face anti-spoofing by using an advanced feature DWT-LBP-DCT and a machine learning classifier SVM. The DWT with a segmentation operation is applied to decompose images into different frequency components, the LBP operation is utilized to extract the spatial information, and the DCT operation is performed to take advantage of energy concentration property to obtain the temporal information by efficiently combining multiple video frames. Finally, the

CRediT authorship contribution statement

Wanling Zhang: Conceptualization, Methodology, Software, Data curation, Writing - original draft. Shijun Xiang: Conceptualization, Methodology, Writing - original draft, Validation.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgments

This work was supported by the National Natural Science Foundation of China (No. 61772234) and the Special Funds for the Cultivation of Guangdong College Students’ Scientific and Technological Innovation, China (No. pdjh2020a0060).

Wanling Zhang was born in JiangXi, China, in 1997. She received the B.S. degree from Jinan University, China, in 2018. She is currently pursuing the Master degree in Jinan University, China. Her current research interests include face anti-spoofing and reversible data hiding.

References (32)

FengL. et al.
Integration of image quality and motion cues for face anti-spoofing
J. Vis. Commun. Image Represent.
(2016)
ZhangJ. et al.
3D Visual passcode: Speech-driven 3D facial dynamics for behaviometrics
Signal Process.
(2019)
WendtH. et al.
Wavelet leaders and bootstrap for multifractal analysis of images
Signal Process.
(2009)
SouzaL. et al.
How far did we get in face spoofing detection?
Eng. Appl. Artif. Intell.
(2017)
LiM. et al.
Personal identification based on iris texture analysis
IEEE Trans. Pattern Anal. Mach. Intell.
(2003)
LeonP.L.D. et al.
Evaluation of speaker verification security and detection of HMM-based synthetic speech
IEEE Trans. Audio Speech Lang. Process.
(2012)
MarascoE. et al.
A survey on antispoofing schemes for fingerprint recognition systems
Acm Comput. Surv.
(2015)
GriechischE. et al.
Online signature verification based on Kolmogorov-Smirnov distribution distance
TianY. et al.
Detection of video-based face spoofing using LBP and multiscale DCT
MäättäJ. et al.
Face spoofing detection from single images using micro-texture analysis

ChingovskaI. et al.

On the effectiveness of local binary patterns in face anti-spoofing

Z. Boulkenafet, J. Komulainen, A. Hadid, Face anti-spoofing based on color texture analysis, in: IEEE International...

MäättäJ. et al.

Face spoofing detection from single images using texture and local shape analysis

Iet Biometr.

(2012)

J. Yang, L. Zhen, S. Liao, S.Z. Li, Face liveness detection with component dependent descriptor, in: International...

Z. Zhang, J. Yan, S. Liu, L. Zhen, S.Z. Li, A face antispoofing database with diverse attacks, in: Iapr International...

B. Peixoto, C. Michelassi, A. Rocha, Face liveness detection under bad illumination conditions, in: IEEE International...

Cited by (24)

Face spoofing detection using Heterogeneous Auto-Similarities of Characteristics
2024, Engineering Applications of Artificial Intelligence
Facial recognition systems are widely used in various daily applications due to their high accuracy rate and ease of use. Unfortunately, their performance drops drastically when encountering spoofing attacks. In this paper, a novel pipeline for face spoofing detection based on the extraction of Heterogeneous Auto-Similarities of Characteristics (HASC) descriptor from the HSV and YCbCr color spaces is proposed. To detect print and replay spoofing attacks, the HASC descriptor encodes linear and non-linear dependencies between low-level dense features of the color face image using covariance and information-theoretic measures. The combination of covariance and information-theoretic measures makes HASC an efficient descriptor to explore how the features are interrelated in real and fake faces. The HASC texture descriptions are extracted separately from each color channel of the HSV and YCbCr color spaces, and the resulting feature vectors are concatenated into an extended feature vector to better discriminate between real and fake faces. Extensive experiments are conducted on publicly available databases namely, Replay-Attack, CASIA FASD, and MSU MFSD. The obtained results demonstrate that while being of low dimension, the HASC descriptor when fed into a binary Support Vector Machine (SVM) classifier, achieves competitive results compared to the state-of-the-art.
Deep learning based intelligent system for robust face spoofing detection using texture feature measurement
2023, Measurement: Sensors
The use of biometric structures in our everyday lives is becoming increasingly frequent. Biometrics play a crucial role in various applications, including crime analysis, person identification, and verification. Among different biometrics, the face provides a rich set of features. However, face spoofing is a continuous threat in real-world environments, leading to abnormal behavior in security systems. Conventional face spoofing analysis methods have often failed to achieve optimal performance in detecting face spoofing, faking, and attacks due to limitations in low-level feature extraction. To address this issue, this research introduces a novel technique called Face Spoofing Detection (FSD) based on Deep Learning Convolutional Neural Network (DLCNN), named NLBP-Net. In this technique, features from face images are extracted using Local Binary Pattern (LBP). The periocular area, which remains untouched by the gradient process, is a section of the human face that stands out as being highly unique. Furthermore, the extracted features are trained using the advanced Visual Geometry Group 16 (VGG16) methods. The trained model effectively classifies spoofing and faking attempts in random face images. Simulations conducted on a standard dataset demonstrate that the proposed NLBP-Net outperforms other methods across several metrics. NLBP-Net achieves an accuracy of 99.593%, sensitivity of 99.633%, specificity of 99.137%, recall of 99.224%, precision of 99.057%, and F1-score of 99.057%.
Face spoofing detection based on multi-scale color inversion dual-stream convolutional neural network
2023, Expert Systems with Applications
Currently, face recognition technology (FRT) has been applied ubiquitously. However, due to the abuse of personal face photos on social media, FRT has encountered unprecedented challenges which promote the development of face spoofing detection (also called face liveness detection or face anti-spoofing) technology. Traditional face spoofing detection methods usually extract features manually and distinguish real and fake faces through a single cue, which may make these methods have problems with low accuracy and generality. In addition, the effectiveness of existing methods is affected by illumination variations. To address the above issues, we propose a multi-scale color inversion dual-stream convolutional neural network, termed MSCI-DSCNN. One stream of the proposed model converts the input RGB images into grayscale ones and conducts multi-scale color inversion to obtain the MSCI images, which are then put into the improved MobileNet to extract face reflection features. The other stream of the network directly feeds RGB images into the improved MobileNet to extract face color features. Finally, the features extracted separately from the two branches are fused and then used for face spoofing detection. We evaluate the proposed framework on three publicly available databases, CASIA-FASD, REPLAY-ATTACK, and OULU-NPU, and achieve promising results. To further measure the generalization capability of the proposed approach, extensive cross-database experiments are performed and the results exhibit great effectiveness of our MSCI-DSCNN method.
Deep malware detection framework for IoT-based smart agriculture
2022, Computers and Electrical Engineering
The advancement in smart agriculture through the Internet of Things (IoT) devices has increased the risk of cyber-attacks. Most of the existing malware detection techniques are unable to detect new variants of malware and suffer from poor accuracy. To overcome the challenges of new malware, this research work proposes a novel three-phase Deep Malware Detection (DMD) framework based on the fusion of Discrete Wavelet Transform (DWT) and Generative Adversarial Network (GAN) named as DMD-DWT-GAN, for IoT-based Smart Agriculture (IoT-SA). This work applied DWT for multiresolution analysis by decomposing the image into Approximation coefficients (Ac) and Detail coefficients (Dc). Finally, a lightweight Convolutional Neural Network (CNN) is employed for in-depth analysis of the malware family. The performance of the proposed framework is evaluated using two benchmark datasets such as IoT malware and Malimg. The proposed framework has achieved 99.99% accuracy on both datasets and is better than the state-of-art models.
Face morphing attack detection and attacker identification based on a watchlist
2022, Signal Processing: Image Communication
Citation Excerpt :
In many countries, facial Automated Border Control (ABC) systems are widely deployed at airports and seaports to improve the efficiency of border crossings [1,2].
Face verification is widely used for Automated Border Control (ABC) in many countries. But such ABC systems are vulnerable to Morphing Attacks (MAs), where a morphed face image is submitted to apply for a passport. To counter face MAs, this paper proposes to detect morphed face images and identify morphing attackers by the use of a watchlist. It is carried out by comparing a suspect image with the biometric references contained in a watchlist, and its detection process is accomplished by analyzing the results of face comparison. Once a morphed image is detected, its morphing attacker is also identified. Meanwhile, a database with different morphing methods, image qualities, facial expressions, and face angles is collected. Experimental results and analysis show that it can achieve stable detection and attacker identification performance for 4 different face MAs, and it can well generalize to unseen morphing types and weights. Moreover, it has good robustness to variations of image qualities, facial expressions, and face angles.
A Review of Face Detection Anti Spoofing Techniques on Varied Data Sets
2024, Lecture Notes in Electrical Engineering

View all citing articles on Scopus

Shijun Xiang received the B.S. degree, the M.S. degree and the Ph.D. degree from Changan University (1997), Guizhou University (2000), and Sun Yatsen University (2006) of China, respectively. From 2006 to 2007, he was a Post-Doctoral Researcher with Korea University, Seoul, South Korea. He is currently a Full Professor with the School of Information Science and Technology, Jinan University, Guangzhou, China. He has authored or co-authored over 60 peer-reviewed papers, including the IEEE TCSVT and the IEEE TMM. His current research interests include robust watermarking, reversible data hiding, secure signal processing in encrypted domain, and face spoofing.

View full text

Face anti-spoofing detection based on DWT-LBP-DCT features

Highlights

Abstract

Introduction

Section snippets

Related work

Proposed method

Face anti-spoofing databases

Experimental results

Conclusion and future work

CRediT authorship contribution statement

Declaration of Competing Interest

Acknowledgments

J. Vis. Commun. Image Represent.

Signal Process.

Signal Process.

How far did we get in face spoofing detection?

Eng. Appl. Artif. Intell.

Personal identification based on iris texture analysis

IEEE Trans. Pattern Anal. Mach. Intell.

Evaluation of speaker verification security and detection of HMM-based synthetic speech

IEEE Trans. Audio Speech Lang. Process.

A survey on antispoofing schemes for fingerprint recognition systems

Acm Comput. Surv.

Online signature verification based on Kolmogorov-Smirnov distribution distance

Detection of video-based face spoofing using LBP and multiscale DCT

Face spoofing detection from single images using micro-texture analysis

On the effectiveness of local binary patterns in face anti-spoofing

Face spoofing detection from single images using texture and local shape analysis

Iet Biometr.