Adversarial robustness in deep neural networks based on variable attributes of the stochastic ensemble model.

Deep neural networks (DNNs) have been shown to be susceptible to critical vulnerabilities when attacked by adversarial samples. This has prompted the development of attack and defense strategies similar to those used in cyberspace security. The dependence of such strategies on attack and defense mechanisms makes the associated algorithms on both sides appear as closely processes, with the defense method being particularly passive in these processes. Inspired by the dynamic defense approach proposed in cyberspace to address endless arm races, this article defines ensemble quantity, network structure, and smoothing parameters as variable ensemble attributes and proposes a stochastic ensemble strategy based on heterogeneous and redundant sub-models. The proposed method introduces the diversity and randomness characteristic of deep neural networks to alter the fixed correspondence gradient between input and output. The unpredictability and diversity of the gradients make it more difficult for attackers to directly implement white-box attacks, helping to address the extreme transferability and vulnerability of ensemble models under white-box attacks. Experimental comparison of ASR-vs.-distortion curves with different attack scenarios under CIFAR10 preliminarily demonstrates the effectiveness of the proposed method that even the highest-capacity attacker cannot easily outperform the attack success rate associated with the ensemble smoothed model, especially for untargeted attacks.

A Mixed Visual Encoding Model Based on the Larger-Scale Receptive Field for Human Brain Activity.

Research on visual encoding models for functional magnetic resonance imaging derived from deep neural networks, especially CNN (e.g., VGG16), has been developed. However, CNNs typically use smaller kernel sizes (e.g., 3 × 3) for feature extraction in visual encoding models. Although the receptive field size of CNN can be enlarged by increasing the network depth or subsampling, it is limited by the small size of the convolution kernel, leading to an insufficient receptive field size. In biological research, the size of the neuronal population receptive field of high-level visual encoding regions is usually three to four times that of low-level visual encoding regions. Thus, CNNs with a larger receptive field size align with the biological findings. The RepLKNet model directly expands the convolution kernel size to obtain a larger-scale receptive field. Therefore, this paper proposes a mixed model to replace CNN for feature extraction in visual encoding models. The proposed model mixes RepLKNet and VGG so that the mixed model has a receptive field of different sizes to extract more feature information from the image. The experimental results indicate that the mixed model achieves better encoding performance in multiple regions of the visual cortex than the traditional convolutional model. Also, a larger-scale receptive field should be considered in building visual encoding models so that the convolution network can play a more significant role in visual representations.

ShapeEditor: A StyleGAN Encoder for Stable and High Fidelity Face Swapping.

With the continuous development of deep-learning technology, ever more advanced face-swapping methods are being proposed. Recently, face-swapping methods based on generative adversarial networks (GANs) have realized many-to-many face exchanges with few samples, which advances the development of this field. However, the images generated by previous GAN-based methods often show instability. The fundamental reason is that the GAN in these frameworks is difficult to converge to the distribution of face space in training completely. To solve this problem, we propose a novel face-swapping method based on pretrained StyleGAN generator with a stronger ability of high-quality face image generation. The critical issue is how to control StyleGAN to generate swapped images accurately. We design the control strategy of the generator based on the idea of encoding and decoding and propose an encoder called ShapeEditor to complete this task. ShapeEditor is a two-step encoder used to generate a set of coding vectors that integrate the identity and attribute of the input faces. In the first step, we extract the identity vector of the source image and the attribute vector of the target image; in the second step, we map the concatenation of the identity vector and attribute vector onto the potential internal space of StyleGAN. Extensive experiments on the test dataset show that the results of the proposed method are not only superior in clarity and authenticity than other state-of-the-art methods but also sufficiently integrate identity and attribute.

Multicenter Computer-Aided Diagnosis for Lymph Nodes Using Unsupervised Domain-Adaptation Networks Based on Cross-Domain Confounding Representations.

To achieve the robust high-performance computer-aided diagnosis systems for lymph nodes, CT images may be typically collected from multicenter data, which cause the isolated performance of the model based on different data source centers. The variability adaptation problem of lymph node data which is related to the problem of domain adaptation in deep learning differs from the general domain adaptation problem because of the typically larger CT image size and more complex data distributions. Therefore, domain adaptation for this problem needs to consider the shared feature representation and even the conditioning information of each domain so that the adaptation network can capture significant discriminative representations in a domain-invariant space. This paper extracts domain-invariant features based on a cross-domain confounding representation and proposes a cycle-consistency learning framework to encourage the network to preserve class-conditioning information through cross-domain image translations. Compared with the performance of different domain adaptation methods, the accurate rate of our method achieves at least 4.4% points higher under multicenter lymph node data. The pixel-level cross-domain image mapping and the semantic-level cycle consistency provided a stable confounding representation with class-conditioning information to achieve effective domain adaptation under complex feature distribution.