论文检索

Social Aspects/Trustworthy Machine Learning

Shaokui Wei, Mingda Zhang, Hongyuan Zha, Baoyuan Wu

Backdoor attacks are serious security threats to machine learning models where an adversary can inject poisoned samples into the training set, causing a backdoored model which predicts poisoned samples with particular triggers to particular target classes, while behaving normally on benign samples. In this paper, we explore the task of purifying a backdoored model using a small clean dataset. By establishing the connection between backdoor risk and adversarial risk, we derive a novel upper bound for backdoor risk, which mainly captures the risk on the shared adversarial examples (SAEs) between the backdoored model and the purified model. This upper bound further suggests a novel bi-level optimization problem for mitigating backdoor using adversarial training techniques. To solve it, we propose Shared Adversarial Unlearning (SAU). Specifically, SAU first generates SAEs, and then, unlearns the generated SAEs such that they are either correctly classified by the purified model and/or differently classified by the two models, such that the backdoor effect in the backdoored model will be mitigated in the purified model. Experiments on various benchmark datasets and network architectures show that our proposed method achieves state-of-the-art performance for backdoor defense. The code is available at https://github.com/SCLBD/BackdoorBench (PyTorch) and https://github.com/shawkui/MindTrojan (MindSpore).

Social Aspects/Trustworthy Machine Learning

Gon Buzaglo, Niv Haim, Gilad Yehudai, Gal Vardi, Yakir Oz, Yaniv Nikankin, Michal Irani

Memorization of training data is an active research area, yet our understanding of the inner workings of neural networks is still in its infancy.Recently, Haim et al. 2022 proposed a scheme to reconstruct training samples from multilayer perceptron binary classifiers, effectively demonstrating that a large portion of training samples are encoded in the parameters of such networks.In this work, we extend their findings in several directions, including reconstruction from multiclass and convolutional neural networks. We derive a more general reconstruction scheme which is applicable to a wider range of loss functions such as regression losses. Moreover, we study the various factors that contribute to networks' susceptibility to such reconstruction schemes. Intriguingly, we observe that using weight decay during training increases reconstructability both in terms of quantity and quality. Additionally, we examine the influence of the number of neurons relative to the number of training samples on the reconstructability.Code: https://github.com/gonbuzaglo/decoreco

Social Aspects/Trustworthy Machine Learning

wenqiang wang, Chongyang Du, Tao Wang, Kaihao Zhang, Wenhan Luo, Lin Ma, Wei Liu, Xiaochun Cao

The adversarial attacks have attracted increasing attention in various fields including natural language processing. The current textual attacking models primarily focus on fooling models by adding character-/word-/sentence-level perturbations, ignoring their influence on human perception. In this paper, for the first time in the community, we propose a novel mode of textual attack, punctuation-level attack. With various types of perturbations, including insertion, displacement, deletion, and replacement, the punctuation-level attack achieves promising fooling rates against SOTA models on typical textual tasks and maintains minimal influence on human perception and understanding of the text by mere perturbation of single-shot single punctuation. Furthermore, we propose a search method named Text Position Punctuation Embedding and Paraphrase (TPPEP) to accelerate the pursuit of optimal position to deploy the attack, without exhaustive search, and we present a mathematical interpretation of TPPEP. Thanks to the integrated Text Position Punctuation Embedding (TPPE), the punctuation attack can be applied at a constant cost of time. Experimental results on public datasets and SOTA models demonstrate the effectiveness of the punctuation attack and the proposed TPPE. We additionally apply the single punctuation attack to summarization, semantic-similarity-scoring, and text-to-image tasks, and achieve encouraging results.

Social Aspects/Trustworthy Machine Learning

Zikai Xiao, Zihan Chen, Songshang Liu, Hualiang Wang, YANG FENG, Jin Hao, Joey Tianyi Zhou, Jian Wu, Howard Yang, Zuozhu Liu

Data privacy and long-tailed distribution are the norms rather than the exception in many real-world tasks. This paper investigates a federated long-tailed learning (Fed-LT) task in which each client holds a locally heterogeneous dataset; if the datasets can be globally aggregated, they jointly exhibit a long-tailed distribution. Under such a setting, existing federated optimization and/or centralized long-tailed learning methods hardly apply due to challenges in (a) characterizing the global long-tailed distribution under privacy constraints and (b) adjusting the local learning strategy to cope with the head-tail imbalance. In response, we propose a method termed $\texttt{Fed-GraB}$, comprised of a Self-adjusting Gradient Balancer (SGB) module that re-weights clients' gradients in a closed-loop manner, based on the feedback of global long-tailed distribution evaluated by a Direct Prior Analyzer (DPA) module. Using $\texttt{Fed-GraB}$, clients can effectively alleviate the distribution drift caused by data heterogeneity during the model training process and obtain a global model with better performance on the minority classes while maintaining the performance of the majority classes. Extensive experiments demonstrate that $\texttt{Fed-GraB}$ achieves state-of-the-art performance on representative datasets such as CIFAR-10-LT, CIFAR-100-LT, ImageNet-LT, and iNaturalist.

Social Aspects/Trustworthy Machine Learning

Xiyuan Yang, Wenke Huang, Mang Ye

Personalized federated learning with differential privacy has been considered a feasible solution to address non-IID distribution of data and privacy leakage risks. However, current personalized federated learning methods suffer from inflexible personalization and convergence difficulties due to two main factors: 1) Firstly, we observe that the prevailing personalization methods mainly achieve this by personalizing a fixed portion of the model, which lacks flexibility. 2) Moreover, we further demonstrate that the default gradient calculation is sensitive to the widely-used clipping operations in differential privacy, resulting in difficulties in convergence. Considering that Fisher information values can serve as an effective measure for estimating the information content of parameters by reflecting the model sensitivity to parameters, we aim to leverage this property to address the aforementioned challenges. In this paper, we propose a novel federated learning method with Dynamic Fisher Personalization and Adaptive Constraint (FedDPA) to handle these challenges. Firstly, by using layer-wise Fisher information to measure the information content of local parameters, we retain local parameters with high Fisher values during the personalization process, which are considered informative, simultaneously prevent these parameters from noise perturbation. Secondly, we introduce an adaptive approach by applying differential constraint strategies to personalized parameters and shared parameters identified in the previous for better convergence. Our method boosts performance through flexible personalization while mitigating the slow convergence caused by clipping operations. Experimental results on CIFAR-10, FEMNIST and SVHN dataset demonstrate the effectiveness of our approach in achieving better performance and robustness against clipping, under personalized federated learning with differential privacy.

Social Aspects/Trustworthy Machine Learning

Siwon Kim, Sangdoo Yun, Hwaran Lee, Martin Gubri, Sungroh Yoon, Seong Joon Oh

The rapid advancement and widespread use of large language models (LLMs) have raised significant concerns regarding the potential leakage of personally identifiable information (PII). These models are often trained on vast quantities of web-collected data, which may inadvertently include sensitive personal data. This paper presents ProPILE, a novel probing tool designed to empower data subjects, or the owners of the PII, with awareness of potential PII leakage in LLM-based services. ProPILE lets data subjects formulate prompts based on their own PII to evaluate the level of privacy intrusion in LLMs. We demonstrate its application on the OPT-1.3B model trained on the publicly available Pile dataset. We show how hypothetical data subjects may assess the likelihood of their PII being included in the Pile dataset being revealed. ProPILE can also be leveraged by LLM service providers to effectively evaluate their own levels of PII leakage with more powerful prompts specifically tuned for their in-house models. This tool represents a pioneering step towards empowering the data subjects for their awareness and control over their own data on the web.

Social Aspects/Trustworthy Machine Learning

Yucheng Shi, Mengnan Du, Xuansheng Wu, Zihan Guan, Jin Sun, Ninghao Liu

Backdoor attacks inject poisoned samples into the training data, resulting in the misclassification of the poisoned input during a model's deployment. Defending against such attacks is challenging, especially for real-world black-box models where only query access is permitted. In this paper, we propose a novel defense framework against backdoor attacks through Zero-shot Image Purification (ZIP). Our framework can be applied to poisoned models without requiring internal information about the model or any prior knowledge of the clean/poisoned samples. Our defense framework involves two steps. First, we apply a linear transformation (e.g., blurring) on the poisoned image to destroy the backdoor pattern. Then, we use a pre-trained diffusion model to recover the missing semantic information removed by the transformation. In particular, we design a new reverse process by using the transformed image to guide the generation of high-fidelity purified images, which works in zero-shot settings. We evaluate our ZIP framework on multiple datasets with different types of attacks. Experimental results demonstrate the superiority of our ZIP framework compared to state-of-the-art backdoor defense baselines. We believe that our results will provide valuable insights for future defense methods for black-box models. Our code is available at https://github.com/sycny/ZIP.

Social Aspects/Trustworthy Machine Learning

Xinyi Xu, Thanh Lam, Chuan Sheng Foo, Bryan Kian Hsiang Low

Valuation methods of data and machine learning (ML) models are essential to the establishment of AI marketplaces. Importantly, certain practical considerations (e.g., operational constraints, legal restrictions) favor the use of model valuation over data valuation. Also, existing marketplaces that involve trading of pre-trained ML models call for an equitable model valuation method to price them. In particular, we investigate the black-box access setting which allows querying a model (to observe predictions) without disclosing model-specific information (e.g., architecture and parameters). By exploiting a Dirichlet abstraction of a model’s predictions, we propose a novel and equitable model valuation method called model Shapley. We also leverage a Lipschitz continuity of model Shapley to design a learning approach for predicting the model Shapley values (MSVs) of many vendors’ models (e.g., 150) in a large-scale marketplace. We perform extensive empirical validation on the effectiveness of model Shapley using various real-world datasets and heterogeneous model types.

Social Aspects/Trustworthy Machine Learning

Zhaoyu Chen, Bo Li, Shuang Wu, Kaixun Jiang, Shouhong Ding, Wenqiang Zhang

Unrestricted adversarial attacks typically manipulate the semantic content of an image (e.g., color or texture) to create adversarial examples that are both effective and photorealistic, demonstrating their ability to deceive human perception and deep neural networks with stealth and success. However, current works usually sacrifice unrestricted degrees and subjectively select some image content to guarantee the photorealism of unrestricted adversarial examples, which limits its attack performance. To ensure the photorealism of adversarial examples and boost attack performance, we propose a novel unrestricted attack framework called Content-based Unrestricted Adversarial Attack. By leveraging a low-dimensional manifold that represents natural images, we map the images onto the manifold and optimize them along its adversarial direction. Therefore, within this framework, we implement Adversarial Content Attack (ACA) based on Stable Diffusion and can generate high transferable unrestricted adversarial examples with various adversarial contents. Extensive experimentation and visualization demonstrate the efficacy of ACA, particularly in surpassing state-of-the-art attacks by an average of 13.3-50.4\% and 16.8-48.0\% in normally trained models and defense methods, respectively.

Social Aspects/Trustworthy Machine Learning

Xun Xian, Ganghua Wang, Jayanth Srinivasa, Ashish Kundu, Xuan Bi, Mingyi Hong, Jie Ding

Backdoor attacks involve inserting poisoned samples during training, resulting in a model containing a hidden backdoor that can trigger specific behaviors without impacting performance on normal samples. These attacks are challenging to detect, as the backdoored model appears normal until activated by the backdoor trigger, rendering them particularly stealthy. In this study, we devise a unified inference-stage detection framework to defend against backdoor attacks. We first rigorously formulate the inference-stage backdoor detection problem, encompassing various existing methods, and discuss several challenges and limitations. We then propose a framework with provable guarantees on the false positive rate or the probability of misclassifying a clean sample. Further, we derive the most powerful detection rule to maximize the detection power, namely the rate of accurately identifying a backdoor sample, given a false positive rate under classical learning scenarios. Based on the theoretically optimal detection rule, we suggest a practical and effective approach for real-world applications based on the latent representations of backdoored deep nets. We extensively evaluate our method on 14 different backdoor attacks using Computer Vision (CV) and Natural Language Processing (NLP) benchmark datasets. The experimental findings align with our theoretical results. We significantly surpass the state-of-the-art methods, e.g., up to 300\% improvement on the detection power as evaluated by AUCROC, over the state-of-the-art defense against advanced adaptive backdoor attacks.

Social Aspects/Trustworthy Machine Learning

Zhen Xiang, Zidi Xiong, Bo Li

Backdoor attack is a common threat to deep neural networks. During testing, samples embedded with a backdoor trigger will be misclassified as an adversarial target by a backdoored model, while samples without the backdoor trigger will be correctly classified. In this paper, we present the first certified backdoor detector (CBD), which is based on a novel, adjustable conformal prediction scheme based on our proposed statistic local dominant probability. For any classifier under inspection, CBD provides 1) a detection inference, 2) the condition under which the attacks are guaranteed to be detectable for the same classification domain, and 3) a probabilistic upper bound for the false positive rate. Our theoretical results show that attacks with triggers that are more resilient to test-time noise and have smaller perturbation magnitudes are more likely to be detected with guarantees. Moreover, we conduct extensive experiments on four benchmark datasets considering various backdoor types, such as BadNet, CB, and Blend. CBD achieves comparable or even higher detection accuracy than state-of-the-art detectors, and it in addition provides detection certification. Notably, for backdoor attacks with random perturbation triggers bounded by $\ell_2\leq0.75$ which achieves more than 90\% attack success rate, CBD achieves 100\% (98\%), 100\% (84\%), 98\% (98\%), and 72\% (40\%) empirical (certified) detection true positive rates on the four benchmark datasets GTSRB, SVHN, CIFAR-10, and TinyImageNet, respectively, with low false positive rates.

Social Aspects/Trustworthy Machine Learning

RENCHUNZI XIE, Hongxin Wei, Lei Feng, Yuzhou Cao, Bo An

Estimating the generalization performance is practically challenging on out-of-distribution (OOD) data without ground-truth labels. While previous methods emphasize the connection between distribution difference and OOD accuracy, we show that a large domain gap not necessarily leads to a low test accuracy. In this paper, we investigate this problem from the perspective of feature separability empirically and theoretically. Specifically, we propose a dataset-level score based upon feature dispersion to estimate the test accuracy under distribution shift. Our method is inspired by desirable properties of features in representation learning: high inter-class dispersion and high intra-class compactness. Our analysis shows that inter-class dispersion is strongly correlated with the model accuracy, while intra-class compactness does not reflect the generalization performance on OOD data. Extensive experiments demonstrate the superiority of our method in both prediction performance and computational efficiency.

Social Aspects/Trustworthy Machine Learning

Francis Ward, Francesca Toni, Francesco Belardinelli, Tom Everitt

Deceptive agents are a challenge for the safety, trustworthiness, and cooperation of AI systems. We focus on the problem that agents might deceive in order to achieve their goals (for instance, in our experiments with language models, the goal of being evaluated as truthful).There are a number of existing definitions of deception in the literature on game theory and symbolic AI, but there is no overarching theory of deception for learning agents in games. We introduce a formaldefinition of deception in structural causal games, grounded in the philosophyliterature, and applicable to real-world machine learning systems.Several examples and results illustrate that our formal definition aligns with the philosophical and commonsense meaning of deception.Our main technical result is to provide graphical criteria for deception. We show, experimentally, that these results can be used to mitigate deception in reinforcement learning agents and language models.

Social Aspects/Trustworthy Machine Learning

Yuzhou Cao, Hussein Mozannar, Lei Feng, Hongxin Wei, Bo An

Enabling machine learning classifiers to defer their decision to a downstream expert when the expert is more accurate will ensure improved safety and performance. This objective can be achieved with the learning-to-defer framework which aims to jointly learn how to classify and how to defer to the expert. In recent studies, it has been theoretically shown that popular estimators for learning to defer parameterized with softmax provide unbounded estimates for the likelihood of deferring which makes them uncalibrated. However, it remains unknown whether this is due to the widely used softmax parameterization and if we can find a softmax-based estimator that is both statistically consistent and possesses a valid probability estimator. In this work, we first show that the cause of the miscalibrated and unbounded estimator in prior literature is due to the symmetric nature of the surrogate losses used and not due to softmax. We then propose a novel statistically consistent asymmetric softmax-based surrogate loss that can produce valid estimates without the issue of unboundedness. We further analyze the non-asymptotic properties of our proposed method and empirically validate its performance and calibration on benchmark datasets.

Social Aspects/Trustworthy Machine Learning

Drago Plecko, Elias Bareinboim

As society transitions towards an AI-based decision-making infrastructure, an ever-increasing number of decisions once under control of humans are now delegated to automated systems. Even though such developments make various parts of society more efficient, a large body of evidence suggests that a great deal of care needs to be taken to make such automated decision-making systems fair and equitable, namely, taking into account sensitive attributes such as gender, race, and religion. In this paper, we study a specific decision-making task called outcome control in which an automated system aims to optimize an outcome variable $Y$ while being fair and equitable. The interest in such a setting ranges from interventions related to criminal justice and welfare, all the way to clinical decision-making and public health. In this paper, we first analyze through causal lenses the notion of benefit, which captures how much a specific individual would benefit from a positive decision, counterfactually speaking, when contrasted with an alternative, negative one. We introduce the notion of benefit fairness, which can be seen as the minimal fairness requirement in decision-making, and develop an algorithm for satisfying it. We then note that the benefit itself may be influenced by the protected attribute, and propose causal tools which can be used to analyze this. Finally, if some of the variations of the protected attribute in the benefit are considered as discriminatory, the notion of benefit fairness may need to be strengthened, which leads us to articulating a notion of causal benefit fairness. Using this notion, we develop a new optimization procedure capable of maximizing $Y$ while ascertaining causal fairness in the decision process.

Social Aspects/Trustworthy Machine Learning

Zhenting Wang, Chen Chen, Yi Zeng, Lingjuan Lyu, Shiqing Ma

Image generation techniques have been gaining increasing attention recently, but concerns have been raised about the potential misuse and intellectual property (IP) infringement associated with image generation models. It is, therefore, necessary to analyze the origin of images by inferring if a specific image was generated by a particular model, i.e., origin attribution. Existing methods only focus on specific types of generative models and require additional procedures during the training phase or generation phase. This makes them unsuitable for pre-trained models that lack these specific operations and may impair generation quality. To address this problem, we first develop an alteration-free and model-agnostic origin attribution method via reverse-engineering on image generation models, i.e., inverting the input of a particular model for a specific image. Given a particular model, we first analyze the differences in the hardness of reverse-engineering tasks for generated samples of the given model and other images. Based on our analysis, we then propose a method that utilizes the reconstruction loss of reverse-engineering to infer the origin. Our proposed method effectively distinguishes between generated images of a specific generative model and other images, i.e., images generated by other models and real images.

Social Aspects/Trustworthy Machine Learning

Dami Choi, Yonadav Shavit, David Duvenaud

It is important that consumers and regulators can verify the provenance of large neural models to evaluate their capabilities and risks. We introduce the concept of a "Proof-of-Training-Data": any protocol that allows a model trainer to convince a Verifier of the training data that produced a set of model weights. Such protocols could verify the amount and kind of data and compute used to train the model, including whether it was trained on specific harmful or beneficial data sources. We explore efficient verification strategies for Proof-of-Training-Data that are compatible with most current large-model training procedures. These include a method for the model-trainer to verifiably pre-commit to a random seed used in training, and a method that exploits models' tendency to temporarily overfit to training data in order to detect whether a given data-point was included in training. We show experimentally that our verification procedures can catch a wide variety of attacks, including all known attacks from the Proof-of-Learning literature.

Social Aspects/Trustworthy Machine Learning

Wei Liu, Jun Wang, Haozhao Wang, Ruixuan Li, Zhiying Deng, YuanKai Zhang, Yang Qiu

Rationalization aims to strengthen the interpretability of NLP models by extracting a subset of human-intelligible pieces of their inputting texts. Conventional works generally employ the maximum mutual information (MMI) criterion to find the rationale that is most indicative of the target label. However, this criterion can be influenced by spurious features that correlate with the causal rationale or the target label. Instead of attempting to rectify the issues of the MMI criterion, we propose a novel criterion to uncover the causal rationale, termed the Minimum Conditional Dependence (MCD) criterion, which is grounded on our finding that the non-causal features and the target label are \emph{d-separated} by the causal rationale. By minimizing the dependence between the non-selected parts of the input and the target label conditioned on the selected rationale candidate, all the causes of the label are compelled to be selected. In this study, we employ a simple and practical measure for dependence, specifically the KL-divergence, to validate our proposed MCD criterion. Empirically, we demonstrate that MCD improves the F1 score by up to 13.7% compared to previous state-of-the-art MMI-based methods.Our code is in an anonymous repository: https://anonymous.4open.science/r/MCD-CE88.

Social Aspects/Trustworthy Machine Learning

Chenxu Zhao, Wei Qian, Rex Ying, Mengdi Huai

With the growing demand for the right to be forgotten, there is an increasing need for machine learning models to forget sensitive data and its impact. To address this, the paradigm of selective forgetting (a.k.a machine unlearning) has been extensively studied, which aims to remove the impact of requested data from a well-trained model without retraining from scratch. Despite its significant success, limited attention has been given to the security vulnerabilities of the unlearning system concerning malicious data update requests. Motivated by this, in this paper, we explore the possibility and feasibility of malicious data update requests during the unlearning process. Specifically, we first propose a new class of malicious selective forgetting attacks, which involves a static scenario where all the malicious data update requests are provided by the adversary at once. Additionally, considering the sequential setting where the data update requests arrive sequentially, we also design a novel framework for sequential forgetting attacks, which is formulated as a stochastic optimal control problem. We also propose novel optimization algorithms that can find the effective malicious data update requests. We perform theoretical analyses for the proposed selective forgetting attacks, and extensive experimental results validate the effectiveness of our proposed selective forgetting attacks. The source code is available in the supplementary material.

Social Aspects/Trustworthy Machine Learning

Yifei Wang, Liangchen Li, Jiansheng Yang, Zhouchen Lin, Yisen Wang

Adversarial Training (AT) has become arguably the state-of-the-art algorithm for extracting robust features. However, researchers recently notice that AT suffers from severe robust overfitting problems, particularly after learning rate (LR) decay. In this paper, we explain this phenomenon by viewing adversarial training as a dynamic minimax game between the model trainer and the attacker. Specifically, we analyze how LR decay breaks the balance between the minimax game by empowering the trainer with a stronger memorization ability, and show such imbalance induces robust overfitting as a result of memorizing non-robust features. We validate this understanding with extensive experiments, and provide a holistic view of robust overfitting from the dynamics of both the two game players. This understanding further inspires us to alleviate robust overfitting by rebalancing the two players by either regularizing the trainer's capacity or improving the attack strength. Experiments show that the proposed ReBalanced Adversarial Training (ReBAT) can attain good robustness and does not suffer from robust overfitting even after very long training. Code is available at https://github.com/PKU-ML/ReBAT.