The exponential growth of encrypted network traffic poses significant challenges for detecting malicious activities online. The scale of emerging malicious traffic is significantly smaller than that of normal traffic, and the imbalanced data distribution poses challenges for detection. However, most existing methods rely on single-category features for classification, which struggle to detect covert malicious traffic behaviors. In this paper, we introduce a novel semi-supervised approach to identify malicious traffic by leveraging multimodal traffic characteristics. By integrating the sequence and topological information inherent in the traffic, we achieve a multifaceted representation of encrypted traffic. We design two independent neural networks to learn the corresponding sequence and topological features from the traffic. This dual-feature extraction enhances the model’s robustness in detecting anomalies within encrypted traffic. The model is trained using a joint strategy that minimizes both the reconstruction error from the autoencoder and the classification loss, allowing it to effectively utilize limited labeled data alongside a large amount of unlabeled data. A confidence-estimation module enhances the classifier’s ability to detect unknown attacks. Finally, our method is evaluated on two benchmark datasets, UNSW-NB15 and CICIDS2017, under various scenarios, including different training set label ratios and the presence of unknown attacks. Our model outperforms other models by 3.49% and 5.69% in F1 score at labeling rates of 1% and 0.1%, respectively.

Pose variant or self-occlusion is one of the open issues which severely degrades the performance of pose-invariant face recognition (PIFR). Existing solutions to PIFR either have undesirable generalization based on challenging pose normalization or are complicated for implement on account of deep neural network. To relieve the impact of ill-pose on PIFR, we have proposed Cross-Pose Generative Adversarial Networks(CP-GAN) to frontalize the profile face with unaltered identity by learning the mapping between the profile and frontal faces in image space. The generator is an encoder-decoder U-net, and generate frontal face image by fusing multiple profile images to achieve a better performance in PIFR. The siamese discriminative network attends to extract the deep representations of the generated frontal face and the ground truth without introducing extra networks in verification and recognition. Besides the implementable architecture, this problem is well alleviated by introducing a combination of adversarial loss for both the generator and the discriminator, symmetry loss, patch-wise loss, and identity loss guiding an identity reserving property of the generated frontal view. Quantitative and qualitative evaluation on both controlled and in-the-wild datasets attest that the solution we proposed to PIFR presents satisfactory perceptual results and transcends state-of-the-art methods on ill-pose face recognition.

Face recognition in unconstraint surveillance is a complicated problem on account of motion blur, expression variations and low resolution. Recent works have demonstrated that patch-attention is strictly more powerful than convolution in recognition models. In this study, we investigate the task of unconstraint surveillance face recognition. First, a Patch-Attention Generative Adversarial Network (PA-GAN) model is devised to aggregate some robust features on behalf of a set of raw surveillance frames, which not only increases the recognition accuracy but also reduces the computational costs of face matching. Second, an improved center loss function combined with abundant unlabeled surveillance faces is utilized to accurately classify the known identities. With the proposed method, the discriminativeness of the face representations is largely enhanced. Finally, the proposed method is verified in two widely used datasets, IJB-A dataset and QMUL-SurvFace dataset to demonstrate the effectiveness. Evaluation of the algorithm performances in comparison with other state-of-the-art methods indicates that the proposed design can achieve competitive accuracy on both the verification and identification protocols.

Face recognition is one of the hottest issues in the field of computer vision and pattern recognition. Deep learning-based recognition models already have more exceptional recognition ability than the human being on open datasets, but still cannot fully undertake the identity recognition task in real scenarios without human assistance. In this paper, we mainly analyze two obstacles, i.e., domain gap and training data shortage. We propose the unpaired Domain Transfer Generative Adversarial Network (DT-GAN) to relieve these two obstacles. We improve the GAN baseline to bridge the domain gap among datasets by generating images conforming to the style of a target domain by learning the mapping between the source domain and target domain. The generator could synthesize face with an arbitrary viewpoint at the same time. The model is trained with a combination of style transfer loss, identity loss, and pose loss, which ensures the successive domain transfer and data augment. We conduct experiments to testify the effectiveness and reasonability of DT-GAN. Experimental results demonstrate the recognition performance is dramatically boosted after domain transfer and data augment.

Although a large number of works have been done to explore efficient face detection in various scenes, practical face detection in unconstrained condition of varying lighting, pose, scale, and occlusion remains a challenging task. The primary limitation of existing solutions is that they are vulnerable to influence from the wild environment. Practical features extraction plays a crucial part in the face detection of low-quality images. Based on the EfficientNet, this paper builds a novel pyramid attention network to integrate multilevel features with rich context messages. Firstly, a context model is exploited to increase the receptive fields at the beginning of the network. Secondly, stacked pyramid feature attention modules and feature fusion simultaneously selectively integrate the contextual information and enable spatial details, thus enhancing the capacity to detect faces on hard images. In addition, hard samples augmentation of the training sets is conducted, which is beneficial for improving the accuracy. A thorough study on ablation verifies the effect of the proposed strategies. Moreover, extensive experiments on Wider Face and FDDB datasets, remarkably pushing the accuracy both of 96.3% (2%↑), demonstrate the performance of the proposed deep face detector which is superior and outperforms most of the preexisting methods. The method presented in this paper can perform the task of face detection in surveillance images well.