Due to the model aging problem, Deep Neural Networks (DNNs) need updates to adjust them to new data distributions. The common practice leverages incremental learning (IL), e.g., Class-based Incremental Learning (CIL) that updates output labels, to update the model with new data and a limited number of old data. This avoids heavyweight training (from scratch) using conventional methods and saves storage space by reducing the number of old data to store. But it also leads to poor performance in fairness. In this paper, we show that CIL suffers both dataset and algorithm bias problems, and existing solutions can only partially solve the problem. We propose a novel framework, CILIATE, that fixes both dataset and algorithm bias in CIL. It features a novel differential analysis guided dataset and training refinement process that identifies unique and important samples overlooked by existing CIL and enforces the model to learn from them. Through this process, CILIATE improves the fairness of CIL by 17.03%, 22.46%, and 31.79% compared to state-of-the-art methods, iCaRL, BiC, and WA, respectively, based on our evaluation on three popular datasets and widely used ResNet models.
Taking full advantage of the excellent performance of StyleGAN, style transfer-based face swapping methods have been extensively investigated recently. However, these studies require separate face segmentation and blending modules for successful face swapping, and the fixed selection of the manipulated latent code in these works is reckless, thus degrading face swapping quality, generalizability, and practicability. This paper proposes a novel and end-to-end integrated framework for high resolution and attribute preservation face swapping via Adaptive Latent Representation Learning. Specifically, we first design a multi-task dual-space face encoder by sharing the underlying feature extraction network to simultaneously complete the facial region perception and face encoding. This encoder enables us to control the face pose and attribute individually, thus enhancing the face swapping quality. Next, we propose an adaptive latent codes swapping module to adaptively learn the mapping between the facial attributes and the latent codes and select effective latent codes for improved retention of facial attributes. Finally, the initial face swapping image generated by StyleGAN2 is blended with the facial region mask generated by our encoder to address the background blur problem. Our framework integrating facial perceiving and blending into the end-to-end training and testing process can achieve high realistic face-swapping on wild faces without segmentation masks. Experimental results demonstrate the superior performance of our approach over state-of-the-art methods.
Machine-Generated Text (MGT) detection, a task that discriminates MGT from Human-Written Text (HWT), plays a crucial role in preventing misuse of text generative models, which excel in mimicking human writing style recently. Latest proposed detectors usually take coarse text sequence as input and output some good results by fine-tune pretrained models with standard cross-entropy loss. However, these methods fail to consider the linguistic aspect of text (e.g., coherence) and sentence-level structures. Moreover, they lack the ability to handle the low-resource problem which could often happen in practice considering the enormous amount of textual data online. In this paper, we present a coherence-based contrastive learning model named CoCo to detect the possible MGT under low-resource scenario. Inspired by the distinctiveness and permanence properties of linguistic feature, we represent text as a coherence graph to capture its entity consistency, which is further encoded by the pretrained model and graph neural network. To tackle the challenges of data limitations, we employ a contrastive learning framework and propose an improved contrastive loss for making full use of hard negative samples in training stage. The experiment results on two public datasets prove our approach outperforms the state-of-art methods significantly.
The security of artificial intelligence (AI) is an important research area towards safe, reliable, and trustworthy AI systems. To accelerate the research on AI security, the Artificial Intelligence Security Competition (AISC) was organized by the Zhongguancun Laboratory, China Industrial Control Systems Cyber Emergency Response Team, Institute for Artificial Intelligence, Tsinghua University, and RealAI as part of the Zhongguancun International Frontier Technology Innovation Competition (https://www.zgc-aisc.com/en). The competition consists of three tracks, including Deepfake Security Competition, Autonomous Driving Security Competition, and Face Recognition Security Competition. This report will introduce the competition rules of these three tracks and the solutions of top-ranking teams in each track.
As in-the-wild data are increasingly involved in the training stage, machine learning applications become more susceptible to data poisoning attacks. Such attacks typically lead to test-time accuracy degradation or controlled misprediction. In this paper, we investigate the third type of exploitation of data poisoning - increasing the risks of privacy leakage of benign training samples. To this end, we demonstrate a set of data poisoning attacks to amplify the membership exposure of the targeted class. We first propose a generic dirty-label attack for supervised classification algorithms. We then propose an optimization-based clean-label attack in the transfer learning scenario, whereby the poisoning samples are correctly labeled and look "natural" to evade human moderation. We extensively evaluate our attacks on computer vision benchmarks. Our results show that the proposed attacks can substantially increase the membership inference precision with minimum overall test-time model performance degradation. To mitigate the potential negative impacts of our attacks, we also investigate feasible countermeasures.
We propose covert beamforming design frameworks for integrated radar sensing and communication (IRSC) systems, where the radar can covertly communicate with legitimate users under the cover of the probing waveforms without being detected by the eavesdropper. Specifically, by jointly designing the target detection beamformer and communication beamformer, we aim to maximize the radar detection mutual information (MI) (or the communication rate) subject to the covert constraint, the communication rate constraint (or the radar detection MI constraint), and the total power constraint. For the perfect eavesdropper's channel state information (CSI) scenario, we transform the covert beamforming design problems into a series of convex subproblems, by exploiting semidefinite relaxation, which can be solved via the bisection search method. Considering the high complexity of iterative optimization, we further propose a single-iterative covert beamformer design scheme based on the zero-forcing criterion. For the imperfect eavesdropper's CSI scenario, we develop a relaxation and restriction method to tackle the robust covert beamforming design problems. Simulation results demonstrate the effectiveness of the proposed covert beamforming schemes for perfect and imperfect CSI scenarios.
Intelligent reflecting surface (IRS) and device-to-device (D2D) communication are two promising technologies for improving transmission reliability between transceivers in communication systems. In this paper, we consider the design of reliable communication between the access point (AP) and actuators for a downlink multiuser multiple-input single-output (MISO) system in the industrial IoT (IIoT) scenario. We propose a two-stage protocol combining IRS with D2D communication so that all actuators can successfully receive the message from AP within a given delay. The superiority of the protocol is that the communication reliability between AP and actuators is doubly augmented by the IRS-aided first-stage transmission and the second-stage D2D transmission. A joint optimization problem of active and passive beamforming is formulated, which aims to maximize the number of actuators with successful decoding. We study the joint beamforming problem for cases where the channel state information (CSI) is perfect and imperfect. For each case, we develop efficient algorithms that include convergence and complexity analysis. Simulation results demonstrate the necessity and role of IRS with a well-optimized reflection matrix, and the D2D network in promoting reliable communication. Moreover, the proposed protocol can enable reliable communication even in the presence of stringent latency requirements and CSI estimation errors.
Backdoor learning is an emerging and important topic of studying the vulnerability of deep neural networks (DNNs). Many pioneering backdoor attack and defense methods are being proposed successively or concurrently, in the status of a rapid arms race. However, we find that the evaluations of new methods are often unthorough to verify their claims and real performance, mainly due to the rapid development, diverse settings, as well as the difficulties of implementation and reproducibility. Without thorough evaluations and comparisons, it is difficult to track the current progress and design the future development roadmap of the literature. To alleviate this dilemma, we build a comprehensive benchmark of backdoor learning, called BackdoorBench. It consists of an extensible modular based codebase (currently including implementations of 8 state-of-the-art (SOTA) attack and 9 SOTA defense algorithms), as well as a standardized protocol of a complete backdoor learning. We also provide comprehensive evaluations of every pair of 8 attacks against 9 defenses, with 5 poisoning ratios, based on 5 models and 4 datasets, thus 8,000 pairs of evaluations in total. We further present analysis from different perspectives about these 8,000 evaluations, studying the effects of attack against defense algorithms, poisoning ratio, model and dataset in backdoor learning. All codes and evaluations of BackdoorBench are publicly available at \url{https://backdoorbench.github.io}.
With Deep Neural Network (DNN) being integrated into a growing number of critical systems with far-reaching impacts on society, there are increasing concerns on their ethical performance, such as fairness. Unfortunately, model fairness and accuracy in many cases are contradictory goals to optimize. To solve this issue, there has been a number of work trying to improve model fairness by using an adversarial game in model level. This approach introduces an adversary that evaluates the fairness of a model besides its prediction accuracy on the main task, and performs joint-optimization to achieve a balanced result. In this paper, we noticed that when performing backward propagation based training, such contradictory phenomenon has shown on individual neuron level. Based on this observation, we propose FairNeuron, a DNN model automatic repairing tool, to mitigate fairness concerns and balance the accuracy-fairness trade-off without introducing another model. It works on detecting neurons with contradictory optimization directions from accuracy and fairness training goals, and achieving a trade-off by selective dropout. Comparing with state-of-the-art methods, our approach is lightweight, making it scalable and more efficient. Our evaluation on 3 datasets shows that FairNeuron can effectively improve all models' fairness while maintaining a stable utility.
People are not always receptive to their voice data being collected and misused. Training the audio intelligence systems needs these data to build useful features, but the cost for getting permissions or purchasing data is very high, which inevitably encourages hackers to collect these voice data without people's awareness. To discourage the hackers from proactively collecting people's voice data, we are the first to propose a clean-label poisoning attack, called WaveFuzz, which can prevent intelligence audio models from building useful features from protected (poisoned) voice data but still preserve the semantic information to the humans. Specifically, WaveFuzz perturbs the voice data to cause Mel Frequency Cepstral Coefficients (MFCC) (typical representations of audio signals) to generate the poisoned frequency features. These poisoned features are then fed to audio prediction models, which degrades the performance of audio intelligence systems. Empirically, we show the efficacy of WaveFuzz by attacking two representative types of intelligent audio systems, i.e., speaker recognition system (SR) and speech command recognition system (SCR). For example, the accuracies of models are declined by $19.78\%$ when only $10\%$ of the poisoned voice data is to fine-tune models, and the accuracies of models declined by $6.07\%$ when only $10\%$ of the training voice data is poisoned. Consequently, WaveFuzz is an effective technique that enables people to fight back to protect their own voice data, which sheds new light on ameliorating privacy issues.