基于真实人脸类别对抗机制的人脸活体检测算法

针对现有人脸活体检测算法在单一数据集内表现良好而在多个数据集间泛化能力较差的问题，提出一种聚焦于真实人脸的活体检测方法。在数据输入阶段，每轮训练会向网络输入所有源域的真实人脸的同时只随机输入其中一个源域的虚假人脸。在特征学习阶段，使用Resnet18网络作为主干网络，对不同残差块的输出特征进行基于注意力机制的加权融合。利用三元组损失和对抗损失对融合后的真实人脸特征进行领域内和领域间的聚合，利用三元组损失对融合后的虚假人脸特征只进行领域内的聚合。在分类阶段，利用交叉熵损失对所有源域的真实人脸和虚假人脸进行分类。所提方法在4个人脸活体检测数据集中进行了实验，实验结果表明所提方法相比其他方法具有更低的识别错误率和更高的鲁棒性。     

基于3D卷积神经网络的活体人脸检测

非法入侵者通过伪装人脸骗取系统认证,给人脸认证系统带来了严重的威胁。因此,活体人脸检测成了人脸认证系统走向实用必须解决的一个重要课题。现有活体人脸检测方法多为基于照片的人脸攻击方面的研究成果,对于基于视频的人脸攻击,效果并不理想。3D卷积神经网络（Convolutional Neural Network,CNN）具有深度学习的特点,能自动学到图像的分布式特征表示;与2D卷积相比,它能学到连续视频帧的动作信息。本文结合3D卷积神经网络的特性,提出利用3D卷积实现视频人脸伪装检测。通过提取3D卷积神经网络最后全连接层学到的时间空间特征,训练SVM（Support Vector Machine）分类器,实现真实人脸和伪装人脸的分类。实验采用两个人脸伪装公开数据库ReplayAttack和CASIA,实现多尺度内部数据库测试和交叉数据库测试。实验结果相对于纹理特征及2D卷积方法有较大提高,可应用于视频人脸攻击的活体人脸检测。      

基于跨域学习的单样本目标检测方法

针对单样本目标检测样本量较少的问题，提出了一种基于跨域学习的方法。该方法从数据增强的角度出发，增加其他域的数据集作为辅助，增强网络学习能力，同时为解决不同域间存在差异的问题，提出了一种基于图片尺度和实例尺度的跨域学习算法，分别对输入的图片特征与检测网络的候选特征增加域分类器模型，用于增强网络对跨域数据的背景和目标的域适应能力。在两个不同的跨域场景进行实验，其中在PASCAL VOC数据集上与目前主流的单样本目标检测算法进行比较，超过目前最好算法2.8个百分点，从而证明了本文方法可以有效提高单样本目标的检测性能。     

基于Adboost与空间支持向量域分类的人脸检测

为了在AdaBoost算法基础上进一步提高人脸检测率,提出首先运用AdaBoost算法对样本进行训练得到T个分类器,然后通过空间支持向量域分类(SSVDC)方法找到T个分类器的超球半径以及球心。同时,为了提高检测速度,首先对彩色图像进行肤色分割,去掉背景以及非肤色区域,然后计算所测样本的对应T个分类器的特征值,并计算其到各个超球球心的距离,并根据其与超球半径的关系来判断是否为人脸。在ORL人脸库、YALE人脸库以及CMU+MIT人脸库中进行实验。实验结果表明:本文算法比AdaBoost算法具有更高的检测速度与检测率,检测率可达到94.4%。     

角度空间三元组损失微调的人脸识别

针对角度Softmax损失强约束存在的问题,提出一种用角度空间三元组损失对角度Softmax损失预训练模型进行微调的算法。算法首先对原来的卷积神经网络结构进行改进,将1×1卷积核与池化层加在不同残差块间,以进行选择更有效的特征。然后用角度空间下的三元组损失对预训练模型进行微调,以降低困难样本的强约束条件。最后在测试时,分别提取原始人脸图像特征和水平翻转的人脸图像特征,对两个特征相加作为最终的人脸特征表达,以丰富人脸特征信息,提高识别率。实验结果表明,在LFW和YTF人脸数据集分别取得了99.25%和94.52%的识别率,在大规模人脸身份识别中,本文提出的方法在仅用单模型和比较小的训练集就能有效地提高人脸识别率。     

面向人脸验证的可迁移对抗样本生成方法

在人脸识别模型的人脸验证任务中,传统的对抗攻击方法无法快速生成真实自然的对抗样本,且对单模型的白盒攻击迁移到其他人脸识别模型上时攻击效果欠佳。该文提出一种基于生成对抗网络的可迁移对抗样本生成方法TAdvFace。TAdvFace采用注意力生成器提高面部特征的提取能力,利用高斯滤波操作提高对抗样本的平滑度,并用自动调整策略调节身份判别损失权重,能够根据不同的人脸图像快速地生成高质量可迁移的对抗样本。实验结果表明,TAdvFace通过单模型的白盒训练,生成的对抗样本能够在多种人脸识别模型和商业API模型上都取得较好的攻击效果,拥有较好的迁移性。     

在线无人监考防作弊新技术探索

为解决教育资源分布不均衡等问题,发展现代网络教育已经成为全球趋势。无人监考是网络教育体系中重要环节之一。为了确保真学真考,本文针对现阶段远程教育中存在的替学替考、假体攻击、夹带抄袭等作弊行为,探讨了活体人脸检测、跨年代人脸识别、夹带行为检测等安全环节的创新方法。上述新技术有助于提高无人监考的安全性,确保远程网络教育落实到人,推动网络教育建设和发展。     

基于网络事件和深度协议分析的入侵检测研究

针对制约NIDS(基于网络的入侵检测系统)的问题,提出了基于网络事件和深度协议分析的入侵检测模型MIDM,实现了对入侵的分析与综合。扩展了ABNF范式形式化定义网络事件,基于所提出模型重新实现了入侵检测系统。实验证明与当前主流NIDS相比,新模型有效降低了误检率和特征库冗余;具有随网络流量和特征库快速增长,而CPU占用率维持低水平增长的特性,能更好地适应高速网络环境;同时还具有一定的特征泛化和检测未知入侵的能力。     

基于Android平台的实时人脸检测与跟踪

针对Android智能手机自带人脸检测功能效率低、错误率高的问题,提出了一种将OpenCV移植到Android平台的方法,在运行Android系统的嵌入式平台中使用改进的AdaBoost算法,并结合OpenCV库来实现实时人脸检测与跟踪。实验取得了高达9505%的人脸检测准确率和5013 ms的平均检测速率,在保证检测速度的同时比Android自带的人脸检测更具高效性和实用性。     

基于自编码器的零样本扣件检测

基于深度学习的扣件检测需要大量人工标注的扣件图像数据集驱动,然而铁路扣件图像中负样本偏少,不均衡的数据集会使得深度学习模型的泛化能力较差,达不到检测扣件状态的效果.针对该问题,本文提出了一种基于自编码器的零样本扣件检测.首先,使用欠完备自编码器、栈式自编码器和卷积自编码器提取扣件正样本图像特征;然后,通过正样本特征向量与基向量的余弦相似度推断出负样本的分布空间;在检测时将各自编码器算法得出的结果利用多数投票法确定样本属性.实验证明,使用本文方法,在只使用正样本训练的情况下,可以有效地检测出扣件图像的负样本,准确率为95.59％,实现了零样本扣件检测.     

Integration of image quality and motion cues for face anti-spoofing: A neural network approach

Many trait-specific countermeasures to face spoofing attacks have been developed for security of face authentication. However, there is no superior face anti-spoofing technique to deal with every kind of spoofing attack in varying scenarios. In order to improve the generalization ability of face anti-spoofing approaches, an extendable multi-cues integration framework for face anti-spoofing using a hierarchical neural network is proposed, which can fuse image quality cues and motion cues for liveness detection. Shearlet is utilized to develop an image quality-based liveness feature. Dense optical flow is utilized to extract motion-based liveness features. A bottleneck feature fusion strategy can integrate different liveness features effectively. The proposed approach was evaluated on three public face anti-spoofing databases. A half total error rate (HTER) of 0% and an equal error rate (EER) of 0% were achieved on both REPLAY-ATTACK database and 3D-MAD database. An EER of 5.83% was achieved on CASIA-FASD database.     

Presentation attack detection based on two-stream vision transformers with self-attention fusion

Aiming at the performance degradation of the existing presentation attack detection methods due to the illumination variation, a two-stream vision transformers framework (TSViT) based on transfer learning in two complementary spaces is proposed in this paper. The face images of RGB color space and multi-scale retinex with color restoration (MSRCR) space are fed to TSViT to learn the distinguishing features of presentation attack detection. To effectively fuse features from two sources (RGB color space images and MSRCR images), a feature fusion method based on self-attention is built, which can effectively capture the complementarity of two features. Experiments and analysis on Oulu-NPU, CASIA-MFSD, and Replay-Attack databases show that it outperforms most existing methods in intra-database testing and achieves good generalization performance in cross-database testing.     

Robust face anti-spoofing with depth information

With the prevalence of face authentication applications, the prevention of malicious attack from fake faces such as photos or videos, i.e., face anti-spoofing, has attracted much attention recently. However, while an increasing number of works on the face anti-spoofing have been reported based on 2D RGB cameras, most of them cannot handle various attacking methods. In this paper we propose a robust representation jointly modeling 2D textual information and depth information for face anti-spoofing. The textual feature is learned from 2D facial image regions using a convolutional neural network (CNN), and the depth representation is extracted from images captured by a Kinect. A face in front of the camera is classified as live if it is categorized as live using both cues. We collected a face anti-spoofing experimental dataset with depth information, and reported extensive experimental results to validate the robustness of the proposed method.     

基于L1范数稀疏距离测度学习的单类分类算法

 已有单类分类算法通常采用欧氏测度描述样本间相似关系,然而欧氏测度有时难以较好地反映一些数据集样本的内在分布结构,为此提出一种用于改善单类分类器描述性能的高维空间单类数据距离测度学习算法,与已有距离测度学习算法相比,该算法只需提供目标类数据,通过引入样本先验分布正则化项和L1范数惩罚的距离测度稀疏性约束,能有效解决高维空间小样本情况下的单类数据距离测度学习问题,并通过采用分块协调下降算法高效的解决距离测度学习的优化问题.学习得到的距离测度能容易地嵌入到单类分类器中,仿真实验结果表明采用学习得到的距离测度能有效改善单类分类器的描述性能,特别能够改善覆盖分类的描述能力,从而使得单类分类器具有更强的推广能力.     

Face presentation attack detection based on chromatic co-occurrence of local binary pattern and ensemble learning

To counter face presentation attacks in face recognition (FR), color texture has been successfully used for face presentation attack detection (PAD) in recent years. However, the existing research does not fully consider the correlation between different color channels as well as the optimization of classification for face PAD. To resolve these limitations, a face PAD scheme based on chromatic co-occurrence of local binary pattern (CCoLBP) and ensemble learning (EL) is proposed in this paper. A color distortion-based face PAD model is first built, and then the chromatic discrepancies between bona fide faces and artefacts are analyzed. After that, CCoLBP is extracted as the feature to characterize these discrepancies. Meanwhile, an EL based classifier is put forward to reduce the effect of class imbalance and to improve the generalization ability. Experimental results and analysis indicate that the proposed scheme can achieve an overall good performance. Moreover, it can achieve significant improvement in the cross-database test, and its computational complexity can meet the requirement of real time applications.     

Multiscale residual gradient attention for face anti-spoofing

In the field of face anti-spoofing (FAS), how to extract the representative features to distinguish between real and spoof faces and train the corresponding deep networks are two vital issues. In this paper, we propose a simple but effective end-to-end FAS model based on an innovative texture extractor and a depth auxiliary supervision mechanism. In the feature extraction stage, we first design the residual gradient convolutions based on the redesigned gradient operators, which are used to extract fine-grained texture features. The extraction of texture features is based on multiple scales by dividing the texture differences between living and spoofing faces into three levels reasonably. Then we construct a multiscale residual gradient attention (MRGA) to obtain representative texture features from multiple levels texture features. By combining the proposed feature extractor MRGA and existing vision transformer (ViT), the MRGA-ViT is proposed to generate related semantics and obtain final classification results. In the training stage, we also propose a local depth auxiliary supervision based on a novel adjacent depth loss, which utilizes the correlation information of adjacent pixels adequately compared with traditional depth loss. The proposed MRGA-ViT model achieves competitive performance in generalization and stability ability, e.g., the ACER(%) values of intra testing on OULU-NPU database are 1.8, 2.6, 1.6 ± 1.2 and 1.9 ± 2.7 respectively, the AUC(%) of cross type testing attains 99.45 ± 0.57, the ACER(%) values of cross dataset testing are 28.1 and 36.7 respectively. Experimental results prove that the proposed model is competitive to other state-of-the-art works on generalization and stability performance.     

Time-series detection of perspiration as a liveness test in fingerprint devices

Fingerprint scanners may be susceptible to spoofing using artificial materials, or in the worst case, dismembered fingers. An anti-spoofing method based on liveness detection has been developed for use in fingerprint scanners. This method quantifies a specific temporal perspiration pattern present in fingerprints acquired from live claimants. The enhanced perspiration detection algorithm presented here improves our previous work by including other fingerprint scanner technologies; using a larger, more diverse data set; and a shorter time window. Several classification methods were tested in order to separate live and spoof fingerprint images. The dataset included fingerprint images from 33 live subjects, 33 spoofs created with dental material and Play-Doh, and fourteen cadaver fingers. Each method had a different performance with respect to each scanner and time window. However, all the classifiers achieved approximately 90% classification rate for all scanners, using the reduced time window and the more comprehensive training and test sets.     

基于上下文信息与注意力特征的欺骗语音检测

随着语音合成和语音转换技术的快速发展，欺骗语音检测方法仍存在欺骗检测准确率低、通用性差等问题。因此，提出一种基于上下文信息与注意力特征的端到端的欺骗检测方法。该方法基于深度残差收缩网络（DRSN），利用双分支上下文信息协调融合模块（DCCM）聚集丰富的上下文信息，融合基于协调时频注意力机制（CTFA）的特征以获得具有上下文信息的跨维度交互特征，从而最大化捕获伪影的潜力。与最佳基线系统相比，在ASVspoof 2019 LA数据集中，所提方法在EER和t-DCF性能指标上分别降低68%和65%；在ASVspoof 2021 LA数据集中，所提方法的EER和t-DCF分别为4.81和0.311 5，分别降低48%和10%。实验结果表明，所提方法能有效提高欺骗语音检测的准确率和泛化能力。