Jingyi Wang

Xiangshan Gao, Xingjun Ma, Jingyi Wang et al.

Federated learning (FL) is a collaborative learning paradigm where participants jointly train a powerful model without sharing their private data. One desirable property for FL is the implementation of the right to be forgotten (RTBF), i.e., a leaving participant has the right to request to delete its private data from the global model. However, unlearning itself may not be enough to implement RTBF unless the unlearning effect can be independently verified, an important aspect that has been overlooked in the current literature. In this paper, we prompt the concept of verifiable federated unlearning, and propose VeriFi, a unified framework integrating federated unlearning and verification that allows systematic analysis of the unlearning and quantification of its effect, with different combinations of multiple unlearning and verification methods. In VeriFi, the leaving participant is granted the right to verify (RTV), that is, the participant notifies the server before leaving, then actively verifies the unlearning effect in the next few communication rounds. The unlearning is done at the server side immediately after receiving the leaving notification, while the verification is done locally by the leaving participant via two steps: marking (injecting carefully-designed markers to fingerprint the leaver) and checking (examining the change of the global model's performance on the markers). Based on VeriFi, we conduct the first systematic and large-scale study for verifiable federated unlearning, considering 7 unlearning methods and 5 verification methods. Particularly, we propose a more efficient and FL-friendly unlearning method, and two more effective and robust non-invasive-verification methods. We extensively evaluate VeriFi on 7 datasets and 4 types of deep learning models. Our analysis establishes important empirical understandings for more trustworthy federated unlearning.

Huizhong Guo, Jinfeng Li, Jingyi Wang et al.

Deep learning-based recommender systems (DRSs) are increasingly and widely deployed in the industry, which brings significant convenience to people's daily life in different ways. However, recommender systems are also shown to suffer from multiple issues,e.g., the echo chamber and the Matthew effect, of which the notation of "fairness" plays a core role.While many fairness notations and corresponding fairness testing approaches have been developed for traditional deep classification models, they are essentially hardly applicable to DRSs. One major difficulty is that there still lacks a systematic understanding and mapping between the existing fairness notations and the diverse testing requirements for deep recommender systems, not to mention further testing or debugging activities. To address the gap, we propose FairRec, a unified framework that supports fairness testing of DRSs from multiple customized perspectives, e.g., model utility, item diversity, item popularity, etc. We also propose a novel, efficient search-based testing approach to tackle the new challenge, i.e., double-ended discrete particle swarm optimization (DPSO) algorithm, to effectively search for hidden fairness issues in the form of certain disadvantaged groups from a vast number of candidate groups. Given the testing report, by adopting a simple re-ranking mitigation strategy on these identified disadvantaged groups, we show that the fairness of DRSs can be significantly improved. We conducted extensive experiments on multiple industry-level DRSs adopted by leading companies. The results confirm that FairRec is effective and efficient in identifying the deeply hidden fairness issues, e.g., achieving 95% testing accuracy with half to 1/8 time.

Mengdi Zhang, Jun Sun, Jingyi Wang et al.

Discrimination has been shown in many machine learning applications, which calls for sufficient fairness testing before their deployment in ethic-relevant domains such as face recognition, medical diagnosis and criminal sentence. Existing fairness testing approaches are mostly designed for identifying individual discrimination, i.e., discrimination against individuals. Yet, as another widely concerning type of discrimination, testing against group discrimination, mostly hidden, is much less studied. To address the gap, in this work, we propose TESTSGD, an interpretable testing approach which systematically identifies and measures hidden (which we call `subtle' group discrimination} of a neural network characterized by conditions over combinations of the sensitive features. Specifically, given a neural network, TESTSGDfirst automatically generates an interpretable rule set which categorizes the input space into two groups exposing the model's group discrimination. Alongside, TESTSGDalso provides an estimated group fairness score based on sampling the input space to measure the degree of the identified subtle group discrimination, which is guaranteed to be accurate up to an error bound. We evaluate TESTSGDon multiple neural network models trained on popular datasets including both structured data and text data. The experiment results show that TESTSGDis effective and efficient in identifying and measuring such subtle group discrimination that has never been revealed before. Furthermore, we show that the testing results of TESTSGDcan guide generation of new samples to mitigate such discrimination through retraining with negligible accuracy drop.

Nico Schiavone, Jingyi Wang, Shuangzhi Li et al.

Active Learning (AL) and Few Shot Learning (FSL) are two label-efficient methods which have achieved excellent results recently. However, most prior arts in both learning paradigms fail to explore the wealth of the vast unlabelled data. In this study, we address this issue in the scenario where the annotation budget is very limited, yet a large amount of unlabelled data for the target task is available. We frame this work in the context of histopathology where labelling is prohibitively expensive. To this end, we introduce an active few shot learning framework, Myriad Active Learning (MAL), including a contrastive-learning encoder, pseudo-label generation, and novel query sample selection in the loop. Specifically, we propose to massage unlabelled data in a self-supervised manner, where the obtained data representations and clustering knowledge form the basis to activate the AL loop. With feedback from the oracle in each AL cycle, the pseudo-labels of the unlabelled data are refined by optimizing a shallow task-specific net on top of the encoder. These updated pseudo-labels serve to inform and improve the active learning query selection process. Furthermore, we introduce a novel recipe to combine existing uncertainty measures and utilize the entire uncertainty list to reduce sample redundancy in AL. Extensive experiments on two public histopathology datasets show that MAL has superior test accuracy, macro F1-score, and label efficiency compared to prior works, and can achieve a comparable test accuracy to a fully supervised algorithm while labelling only 5% of the dataset.

Jialuo Chen, Jingyi Wang, Xiyue Zhang et al.

Due to the vast testing space, the increasing demand for effective and efficient testing of deep neural networks (DNNs) has led to the development of various DNN test case prioritization techniques. However, the fact that DNNs can deliver high-confidence predictions for incorrectly predicted examples, known as the over-confidence problem, causes these methods to fail to reveal high-confidence errors. To address this limitation, in this work, we propose FAST, a method that boosts existing prioritization methods through guided FeAture SelecTion. FAST is based on the insight that certain features may introduce noise that affects the model's output confidence, thereby contributing to high-confidence errors. It quantifies the importance of each feature for the model's correct predictions, and then dynamically prunes the information from the noisy features during inference to derive a new probability vector for the uncertainty estimation. With the help of FAST, the high-confidence errors and correctly classified examples become more distinguishable, resulting in higher APFD (Average Percentage of Fault Detection) values for test prioritization, and higher generalization ability for model enhancement. We conduct extensive experiments to evaluate FAST across a diverse set of model structures on multiple benchmark datasets to validate the effectiveness, efficiency, and scalability of FAST compared to the state-of-the-art prioritization techniques.

Xiaohan Yuan, Jinfeng Li, Dongxia Wang et al.

Generative large language models (LLMs) have revolutionized natural language processing with their transformative and emergent capabilities. However, recent evidence indicates that LLMs can produce harmful content that violates social norms, raising significant concerns regarding the safety and ethical ramifications of deploying these advanced models. Thus, it is both critical and imperative to perform a rigorous and comprehensive safety evaluation of LLMs before deployment. Despite this need, owing to the extensiveness of LLM generation space, it still lacks a unified and standardized risk taxonomy to systematically reflect the LLM content safety, as well as automated safety assessment techniques to explore the potential risk efficiently. To bridge the striking gap, we propose S-Eval, a novel LLM-based automated Safety Evaluation framework with a newly defined comprehensive risk taxonomy. S-Eval incorporates two key components, i.e., an expert testing LLM ${M}_t$ and a novel safety critique LLM ${M}_c$. ${M}_t$ is responsible for automatically generating test cases in accordance with the proposed risk taxonomy. ${M}_c$ can provide quantitative and explainable safety evaluations for better risk awareness of LLMs. In contrast to prior works, S-Eval is efficient and effective in test generation and safety evaluation. Moreover, S-Eval can be flexibly configured and adapted to the rapid evolution of LLMs and accompanying new safety threats, test generation methods and safety critique methods thanks to the LLM-based architecture. S-Eval has been deployed in our industrial partner for the automated safety evaluation of multiple LLMs serving millions of users, demonstrating its effectiveness in real-world scenarios. Our benchmark is publicly available at https://github.com/IS2Lab/S-Eval.

Yin Luo, Qingchao Kong, Nan Xu et al.

As the latest advancements in natural language processing, large language models (LLMs) have achieved human-level language understanding and generation abilities in many real-world tasks, and even have been regarded as a potential path to the artificial general intelligence. To better facilitate research on LLMs, many open-source LLMs, such as Llama 2 and Falcon, have recently been proposed and gained comparable performances to proprietary models. However, these models are primarily designed for English scenarios and exhibit poor performances in Chinese contexts. In this technical report, we propose YAYI 2, including both base and chat models, with 30 billion parameters. YAYI 2 is pre-trained from scratch on a multilingual corpus which contains 2.65 trillion tokens filtered by our pre-training data processing pipeline. The base model is aligned with human values through supervised fine-tuning with millions of instructions and reinforcement learning from human feedback. Extensive experiments on multiple benchmarks, such as MMLU and CMMLU, consistently demonstrate that the proposed YAYI 2 outperforms other similar sized open-source models.

Xiaokun Luan, Xiyue Zhang, Jingyi Wang et al.

Model reuse techniques can reduce the resource requirements for training high-performance deep neural networks (DNNs) by leveraging existing models. However, unauthorized reuse and replication of DNNs can lead to copyright infringement and economic loss to the model owner. This underscores the need to analyze the reuse relation between DNNs and develop copyright protection techniques to safeguard intellectual property rights. Existing white-box testing-based approaches cannot address the common heterogeneous reuse case where the model architecture is changed, and DNN fingerprinting approaches heavily rely on generating adversarial examples with good transferability, which is known to be challenging in the black-box setting. To bridge the gap, we propose NFARD, a Neuron Functionality Analysis-based Reuse Detector, which only requires normal test samples to detect reuse relations by measuring the models' differences on a newly proposed model characterization, i.e., neuron functionality (NF). A set of NF-based distance metrics is designed to make NFARD applicable to both white-box and black-box settings. Moreover, we devise a linear transformation method to handle heterogeneous reuse cases by constructing the optimal projection matrix for dimension consistency, significantly extending the application scope of NFARD. To the best of our knowledge, this is the first adversarial example-free method that exploits neuron functionality for DNN copyright protection. As a side contribution, we constructed a reuse detection benchmark named Reuse Zoo that covers various practical reuse techniques and popular datasets. Extensive evaluations on this comprehensive benchmark show that NFARD achieves F1 scores of 0.984 and 1.0 for detecting reuse relationships in black-box and white-box settings, respectively, while generating test suites 2 ~ 99 times faster than previous methods.

Tengjin Weng, Jingyi Wang, Wenhao Jiang et al.

Can Multimodal Large Language Models (MLLMs) develop an intuitive number sense similar to humans? Targeting this problem, we introduce Visual Number Benchmark (VisNumBench) to evaluate the number sense abilities of MLLMs across a wide range of visual numerical tasks. VisNumBench consists of about 1,900 multiple-choice question-answer pairs derived from both synthetic and real-world visual data, covering seven visual numerical attributes and four types of visual numerical estimation tasks. Our experiments on VisNumBench led to the following key findings: (i) The 17 MLLMs we tested, including open-source models such as Qwen2.5-VL and InternVL2.5, as well as proprietary models like GPT-4o and Gemini 2.0 Flash, perform significantly below human levels in number sense-related tasks. (ii) Multimodal mathematical models and multimodal chain-of-thought (CoT) models did not exhibit significant improvements in number sense abilities. (iii) Stronger MLLMs with larger parameter sizes and broader general abilities demonstrate modest gains in number sense abilities. We believe VisNumBench will serve as a valuable resource for the research community, encouraging further advancements in enhancing MLLMs' number sense abilities. Code and dataset are available at https://wwwtttjjj.github.io/VisNumBench/.

Xiangxiang Chen, Peixin Zhang, Jun Sun et al.

Model quantization is a popular technique for deploying deep learning models on resource-constrained environments. However, it may also introduce previously overlooked security risks. In this work, we present QuRA, a novel backdoor attack that exploits model quantization to embed malicious behaviors. Unlike conventional backdoor attacks relying on training data poisoning or model training manipulation, QuRA solely works using the quantization operations. In particular, QuRA first employs a novel weight selection strategy to identify critical weights that influence the backdoor target (with the goal of perserving the model's overall performance in mind). Then, by optimizing the rounding direction of these weights, we amplify the backdoor effect across model layers without degrading accuracy. Extensive experiments demonstrate that QuRA achieves nearly 100% attack success rates in most cases, with negligible performance degradation. Furthermore, we show that QuRA can adapt to bypass existing backdoor defenses, underscoring its threat potential. Our findings highlight critical vulnerability in widely used model quantization process, emphasizing the need for more robust security measures. Our implementation is available at https://github.com/cxx122/QuRA.

Botao Ren, Botian Xu, Tengyu Liu et al.

Neuroscience studies have shown that the human visual system utilizes high-level feedback information to guide lower-level perception, enabling adaptation to signals of different characteristics. In light of this, we propose Feedback multi-Level feature Extractor (Flex) to incorporate a similar mechanism for object detection. Flex refines feature selection based on image-wise and instance-level feedback information in response to image quality variation and classification uncertainty. Experimental results show that Flex offers consistent improvement to a range of existing SOTA methods on the challenging aerial object detection datasets including DOTA-v1.0, DOTA-v1.5, and HRSC2016. Although the design originates in aerial image detection, further experiments on MS COCO also reveal our module's efficacy in general detection models. Quantitative and qualitative analyses indicate that the improvements are closely related to image qualities, which match our motivation.

Jingyi Wang, Xiaobo Xia, Long Lan et al.

Given data with noisy labels, over-parameterized deep networks suffer overfitting mislabeled data, resulting in poor generalization. The memorization effect of deep networks shows that although the networks have the ability to memorize all noisy data, they would first memorize clean training data, and then gradually memorize mislabeled training data. A simple and effective method that exploits the memorization effect to combat noisy labels is early stopping. However, early stopping cannot distinguish the memorization of clean data and mislabeled data, resulting in the network still inevitably overfitting mislabeled data in the early training stage.In this paper, to decouple the memorization of clean data and mislabeled data, and further reduce the side effect of mislabeled data, we perform additive decomposition on network parameters. Namely, all parameters are additively decomposed into two groups, i.e., parameters $\mathbf{w}$ are decomposed as $\mathbf{w}=\bmσ+\bmγ$. Afterward, the parameters $\bmσ$ are considered to memorize clean data, while the parameters $\bmγ$ are considered to memorize mislabeled data. Benefiting from the memorization effect, the updates of the parameters $\bmσ$ are encouraged to fully memorize clean data in early training, and then discouraged with the increase of training epochs to reduce interference of mislabeled data. The updates of the parameters $\bmγ$ are the opposite. In testing, only the parameters $\bmσ$ are employed to enhance generalization. Extensive experiments on both simulated and real-world benchmarks confirm the superior performance of our method.

Yao Lu, Peixin Zhang, Jingyi Wang et al.

Deep learning has revolutionized computing in many real-world applications, arguably due to its remarkable performance and extreme convenience as an end-to-end solution. However, deep learning models can be costly to train and to use, especially for those large-scale models, making it necessary to optimize the original overly complicated models into smaller ones in scenarios with limited resources such as mobile applications or simply for resource saving. The key question in such model optimization is, how can we effectively identify and measure the redundancy in a deep learning model structure. While several common metrics exist in the popular model optimization techniques to measure the performance of models after optimization, they are not able to quantitatively inform the degree of remaining redundancy. To address the problem, we present a novel testing approach, i.e., RedTest, which proposes a novel testing metric called Model Structural Redundancy Score (MSRS) to quantitatively measure the degree of redundancy in a deep learning model structure. We first show that MSRS is effective in both revealing and assessing the redundancy issues in many state-of-the-art models, which urgently calls for model optimization. Then, we utilize MSRS to assist deep learning model developers in two practical application scenarios: 1) in Neural Architecture Search, we design a novel redundancy-aware algorithm to guide the search for the optimal model structure and demonstrate its effectiveness by comparing it to existing standard NAS practice; 2) in the pruning of large-scale pre-trained models, we prune the redundant layers of pre-trained models with the guidance of layer similarity to derive less redundant ones of much smaller size. Extensive experimental results demonstrate that removing such redundancy has a negligible effect on the model utility.

Jingyi Wang, Ying Chen, Paolo Giudici

We propose Group Shapley, a metric that extends the classical individual-level Shapley value framework to evaluate the importance of feature groups, addressing the structured nature of predictors commonly found in business and economic data. More importantly, we develop a significance testing procedure based on a three-cumulant chi-square approximation and establish the asymptotic properties of the test statistics for Group Shapley values. Our approach can effectively handle challenging scenarios, including sparse or skewed distributions and small sample sizes, outperforming alternative tests such as the Wald test. Simulations confirm that the proposed test maintains robust empirical size and demonstrates enhanced power under diverse conditions. To illustrate the method's practical relevance in advancing Explainable AI, we apply our framework to bond recovery rate predictions using a global dataset (1996-2023) comprising 2,094 observations and 98 features, grouped into 16 subgroups and five broader categories: bond characteristics, firm fundamentals, industry-specific factors, market-related variables, and macroeconomic indicators. Our results identify the market-related variables group as the most influential. Furthermore, Lorenz curves and Gini indices reveal that Group Shapley assigns feature importance more equitably compared to individual Shapley values.

Xuran Li, Jingyi Wang, Xiaohan Yuan et al.

It is often desirable to remove (a.k.a. unlearn) a specific part of the training data from a trained neural network model. A typical application scenario is to protect the data holder's right to be forgotten, which has been promoted by many recent regulation rules. Existing unlearning methods involve training alternative models with remaining data, which may be costly and challenging to verify from the data holder or a thirdparty auditor's perspective. In this work, we provide a new angle and propose a novel unlearning approach by imposing carefully crafted "patch" on the original neural network to achieve targeted "forgetting" of the requested data to delete. Specifically, inspired by the research line of neural network repair, we propose to strategically seek a lightweight minimum "patch" for unlearning a given data point with certifiable guarantee. Furthermore, to unlearn a considerable amount of data points (or an entire class), we propose to iteratively select a small subset of representative data points to unlearn, which achieves the effect of unlearning the whole set. Extensive experiments on multiple categorical datasets demonstrates our approach's effectiveness, achieving measurable unlearning while preserving the model's performance and being competitive in efficiency and memory consumption compared to various baseline methods.

Guoliang Dong, Jingyi Wang, Jun Sun et al.

It is known that neural networks are subject to attacks through adversarial perturbations, i.e., inputs which are maliciously crafted through perturbations to induce wrong predictions. Furthermore, such attacks are impossible to eliminate, i.e., the adversarial perturbation is still possible after applying mitigation methods such as adversarial training. Multiple approaches have been developed to detect and reject such adversarial inputs, mostly in the image domain. Rejecting suspicious inputs however may not be always feasible or ideal. First, normal inputs may be rejected due to false alarms generated by the detection algorithm. Second, denial-of-service attacks may be conducted by feeding such systems with adversarial inputs. To address the gap, in this work, we propose an approach to automatically repair adversarial texts at runtime. Given a text which is suspected to be adversarial, we novelly apply multiple adversarial perturbation methods in a positive way to identify a repair, i.e., a slightly mutated but semantically equivalent text that the neural network correctly classifies. Our approach has been experimented with multiple models trained for natural language processing tasks and the results show that our approach is effective, i.e., it successfully repairs about 80\% of the adversarial texts. Furthermore, depending on the applied perturbation method, an adversarial text could be repaired in as short as one second on average.

Shunkai Zhu, Jingyi Wang, Jun Sun et al.

Fuzzing is one of the prevailing methods for vulnerability detection. However, even state-of-the-art fuzzing methods become ineffective after some period of time, i.e., the coverage hardly improves as existing methods are ineffective to focus the attention of fuzzing on covering the hard-to-trigger program paths. In other words, they cannot generate inputs that can break the bottleneck due to the fundamental difficulty in capturing the complex relations between the test inputs and program coverage. In particular, existing fuzzers suffer from the following main limitations: 1) lacking an overall analysis of the program to identify the most "rewarding" seeds, and 2) lacking an effective mutation strategy which could continuously select and mutates the more relevant "bytes" of the seeds. In this work, we propose an approach called ATTuzz to address these two issues systematically. First, we propose a lightweight dynamic analysis technique which estimates the "reward" of covering each basic block and selects the most rewarding seeds accordingly. Second, we mutate the selected seeds according to a neural network model which predicts whether a certain "rewarding" block will be covered given certain mutation on certain bytes of a seed. The model is a deep learning model equipped with attention mechanism which is learned and updated periodically whilst fuzzing. Our evaluation shows that ATTuzz significantly outperforms 5 state-of-the-art grey-box fuzzers on 13 popular real-world programs at achieving higher edge coverage and finding new bugs. In particular, ATTuzz achieved 2X edge coverage and 4X bugs detected than AFL over 24-hour runs. Moreover, ATTuzz persistently improves the edge coverage in the long run, i.e., achieving 50% more coverage than AFL in 5 days.

Jialuo Chen, Jingyi Wang, Tinglan Peng et al.

Deep learning (DL) models, especially those large-scale and high-performance ones, can be very costly to train, demanding a great amount of data and computational resources. Unauthorized reproduction of DL models can lead to copyright infringement and cause huge economic losses to model owners. Existing copyright protection techniques are mostly based on watermarking, which embeds an owner-specified watermark into the model. While being able to provide exact ownership verification, these techniques are 1) invasive, as they need to tamper with the training process, which may affect the utility or introduce new security risks; 2) prone to adaptive attacks that attempt to remove the watermark; and 3) not robust to the emerging model extraction attacks. Latest fingerprinting work, though being non-invasive, also falls short when facing the diverse and ever-growing attack scenarios. In this paper, we propose a novel testing framework for DL copyright protection: DEEPJUDGE. DEEPJUDGE quantitatively tests the similarities between two DL models: a victim model and a suspect model. It leverages a diverse set of testing metrics and test case generation methods to produce a chain of supporting evidence to help determine whether a suspect model is a copy of the victim model. Advantages of DEEPJUDGE include: 1) non-invasive, as it works directly on the model and does not tamper with the training process; 2) efficient, as it only needs a small set of test cases and a quick scan of models; 3) flexible, as it can easily incorporate new metrics or generation methods to obtain more confident judgement; and 4) fairly robust to model extraction and adaptive attacks. We verify the effectiveness of DEEPJUDGE under typical copyright infringement scenarios, including model finetuning, pruning and extraction, via extensive experiments on both image and speech datasets with a variety of model architectures.

Peixin Zhang, Jingyi Wang, Jun Sun et al.

As deep image classification applications, e.g., face recognition, become increasingly prevalent in our daily lives, their fairness issues raise more and more concern. It is thus crucial to comprehensively test the fairness of these applications before deployment. Existing fairness testing methods suffer from the following limitations: 1) applicability, i.e., they are only applicable for structured data or text without handling the high-dimensional and abstract domain sampling in the semantic level for image classification applications; 2) functionality, i.e., they generate unfair samples without providing testing criterion to characterize the model's fairness adequacy. To fill the gap, we propose DeepFAIT, a systematic fairness testing framework specifically designed for deep image classification applications. DeepFAIT consists of several important components enabling effective fairness testing of deep image classification applications: 1) a neuron selection strategy to identify the fairness-related neurons; 2) a set of multi-granularity adequacy metrics to evaluate the model's fairness; 3) a test selection algorithm for fixing the fairness issues efficiently. We have conducted experiments on widely adopted large-scale face recognition applications, i.e., VGGFace and FairFace. The experimental results confirm that our approach can effectively identify the fairness-related neurons, characterize the model's fairness, and select the most valuable test cases to mitigate the model's fairness issues.

Peixin Zhang, Jingyi Wang, Jun Sun et al.

Although deep learning has demonstrated astonishing performance in many applications, there are still concerns about its dependability. One desirable property of deep learning applications with societal impact is fairness (i.e., non-discrimination). Unfortunately, discrimination might be intrinsically embedded into the models due to the discrimination in the training data. As a countermeasure, fairness testing systemically identifies discriminatory samples, which can be used to retrain the model and improve the model's fairness. Existing fairness testing approaches however have two major limitations. Firstly, they only work well on traditional machine learning models and have poor performance (e.g., effectiveness and efficiency) on deep learning models. Secondly, they only work on simple structured (e.g., tabular) data and are not applicable for domains such as text. In this work, we bridge the gap by proposing a scalable and effective approach for systematically searching for discriminatory samples while extending existing fairness testing approaches to address a more challenging domain, i.e., text classification. Compared with state-of-the-art methods, our approach only employs lightweight procedures like gradient computation and clustering, which is significantly more scalable and effective. Experimental results show that on average, our approach explores the search space much more effectively (9.62 and 2.38 times more than the state-of-the-art methods respectively on tabular and text datasets) and generates much more discriminatory samples (24.95 and 2.68 times) within a same reasonable time. Moreover, the retrained models reduce discrimination by 57.2% and 60.2% respectively on average.

Zhe Zhao, Guangke Chen, Jingyi Wang et al.

As a new programming paradigm, deep learning has expanded its application to many real-world problems. At the same time, deep learning based software are found to be vulnerable to adversarial attacks. Though various defense mechanisms have been proposed to improve robustness of deep learning software, many of them are ineffective against adaptive attacks. In this work, we propose a novel characterization to distinguish adversarial examples from benign ones based on the observation that adversarial examples are significantly less robust than benign ones. As existing robustness measurement does not scale to large networks, we propose a novel defense framework, named attack as defense (A2D), to detect adversarial examples by effectively evaluating an example's robustness. A2D uses the cost of attacking an input for robustness evaluation and identifies those less robust examples as adversarial since less robust examples are easier to attack. Extensive experiment results on MNIST, CIFAR10 and ImageNet show that A2D is more effective than recent promising approaches. We also evaluate our defence against potential adaptive attacks and show that A2D is effective in defending carefully designed adaptive attacks, e.g., the attack success rate drops to 0% on CIFAR10.

Jingyi Wang, Zhidong Deng

Graph convolutional neural network provides good solutions for node classification and other tasks with non-Euclidean data. There are several graph convolutional models that attempt to develop deep networks but do not cause serious over-smoothing at the same time. Considering that the wavelet transform generally has a stronger ability to extract useful information than the Fourier transform, we propose a new deep graph wavelet convolutional network (DeepGWC) for semi-supervised node classification tasks. Based on the optimized static filtering matrix parameters of vanilla graph wavelet neural networks and the combination of Fourier bases and wavelet ones, DeepGWC is constructed together with the reuse of residual connection and identity mappings in network architectures. Extensive experiments on three benchmark datasets including Cora, Citeseer, and Pubmed are conducted. The experimental results demonstrate that our DeepGWC outperforms existing graph deep models with the help of additional wavelet bases and achieves new state-of-the-art performances eventually.

Jingyi Wang, Jialuo Chen, Youcheng Sun et al.

Recently, there has been a significant growth of interest in applying software engineering techniques for the quality assurance of deep learning (DL) systems. One popular direction is deep learning testing, where adversarial examples (a.k.a.~bugs) of DL systems are found either by fuzzing or guided search with the help of certain testing metrics. However, recent studies have revealed that the commonly used neuron coverage metrics by existing DL testing approaches are not correlated to model robustness. It is also not an effective measurement on the confidence of the model robustness after testing. In this work, we address this gap by proposing a novel testing framework called Robustness-Oriented Testing (RobOT). A key part of RobOT is a quantitative measurement on 1) the value of each test case in improving model robustness (often via retraining), and 2) the convergence quality of the model robustness improvement. RobOT utilizes the proposed metric to automatically generate test cases valuable for improving model robustness. The proposed metric is also a strong indicator on how well robustness improvement has converged through testing. Experiments on multiple benchmark datasets confirm the effectiveness and efficiency of RobOT in improving DL model robustness, with 67.02% increase on the adversarial robustness that is 50.65% higher than the state-of-the-art work DeepGini.

Guoliang Dong, Jun Sun, Jingyi Wang et al.

Neural networks are increasingly applied to support decision making in safety-critical applications (like autonomous cars, unmanned aerial vehicles and face recognition based authentication). While many impressive static verification techniques have been proposed to tackle the correctness problem of neural networks, it is possible that static verification may never be sufficiently scalable to handle real-world neural networks. In this work, we propose a runtime verification method to ensure the correctness of neural networks. Given a neural network and a desirable safety property, we adopt state-of-the-art static verification techniques to identify strategically locations to introduce additional gates which "correct" neural network behaviors at runtime. Experiment results show that our approach effectively generates neural networks which are guaranteed to satisfy the properties, whilst being consistent with the original neural network most of the time.

Pengfei Yang, Renjue Li, Jianlin Li et al.

We propose a spurious region guided refinement approach for robustness verification of deep neural networks. Our method starts with applying the DeepPoly abstract domain to analyze the network. If the robustness property cannot be verified, the result is inconclusive. Due to the over-approximation, the computed region in the abstraction may be spurious in the sense that it does not contain any true counterexample. Our goal is to identify such spurious regions and use them to guide the abstraction refinement. The core idea is to make use of the obtained constraints of the abstraction to infer new bounds for the neurons. This is achieved by linear programming techniques. With the new bounds, we iteratively apply DeepPoly, aiming to eliminate spurious regions. We have implemented our approach in a prototypical tool DeepSRGR. Experimental results show that a large amount of regions can be identified as spurious, and as a result, the precision of DeepPoly can be significantly improved. As a side contribution, we show that our approach can be applied to verify quantitative robustness properties.

Yizhen Dong, Peixin Zhang, Jingyi Wang et al.

Deep neural networks (DNN) are increasingly applied in safety-critical systems, e.g., for face recognition, autonomous car control and malware detection. It is also shown that DNNs are subject to attacks such as adversarial perturbation and thus must be properly tested. Many coverage criteria for DNN since have been proposed, inspired by the success of code coverage criteria for software programs. The expectation is that if a DNN is a well tested (and retrained) according to such coverage criteria, it is more likely to be robust. In this work, we conduct an empirical study to evaluate the relationship between coverage, robustness and attack/defense metrics for DNN. Our study is the largest to date and systematically done based on 100 DNN models and 25 metrics. One of our findings is that there is limited correlation between coverage and robustness, i.e., improving coverage does not help improve the robustness. Our dataset and implementation have been made available to serve as a benchmark for future studies on testing DNN.

Guoliang Dong, Jingyi Wang, Jun Sun et al.

Neural networks are becoming a popular tool for solving many real-world problems such as object recognition and machine translation, thanks to its exceptional performance as an end-to-end solution. However, neural networks are complex black-box models, which hinders humans from interpreting and consequently trusting them in making critical decisions. Towards interpreting neural networks, several approaches have been proposed to extract simple deterministic models from neural networks. The results are not encouraging (e.g., low accuracy and limited scalability), fundamentally due to the limited expressiveness of such simple models. In this work, we propose an approach to extract probabilistic automata for interpreting an important class of neural networks, i.e., recurrent neural networks. Our work distinguishes itself from existing approaches in two important ways. One is that probability is used to compensate for the loss of expressiveness. This is inspired by the observation that human reasoning is often `probabilistic'. The other is that we adaptively identify the right level of abstraction so that a simple model is extracted in a request-specific way. We conduct experiments on several real-world datasets using state-of-the-art architectures including GRU and LSTM. The result shows that our approach significantly improves existing approaches in terms of accuracy or scalability. Lastly, we demonstrate the usefulness of the extracted models through detecting adversarial texts.

Jingyi Wang, Jun Sun, Peixin Zhang et al.

Recently, it has been shown that deep neural networks (DNN) are subject to attacks through adversarial samples. Adversarial samples are often crafted through adversarial perturbation, i.e., manipulating the original sample with minor modifications so that the DNN model labels the sample incorrectly. Given that it is almost impossible to train perfect DNN, adversarial samples are shown to be easy to generate. As DNN are increasingly used in safety-critical systems like autonomous cars, it is crucial to develop techniques for defending such attacks. Existing defense mechanisms which aim to make adversarial perturbation challenging have been shown to be ineffective. In this work, we propose an alternative approach. We first observe that adversarial samples are much more sensitive to perturbations than normal samples. That is, if we impose random perturbations on a normal and an adversarial sample respectively, there is a significant difference between the ratio of label change due to the perturbations. Observing this, we design a statistical adversary detection algorithm called nMutant (inspired by mutation testing from software engineering community). Our experiments show that nMutant effectively detects most of the adversarial samples generated by recently proposed attacking methods. Furthermore, we provide an error bound with certain statistical significance along with the detection.

Jingyi Wang, Jun Sun, Shengchao Qin et al.

Precisely modeling complex systems like cyber-physical systems is challenging, which often render model-based system verification techniques like model checking infeasible. To overcome this challenge, we propose a method called LAR to automatically `verify' such complex systems through a combination of learning, abstraction and refinement from a set of system log traces. We assume that log traces and sampling frequency are adequate to capture `enough' behaviour of the system. Given a safety property and the concrete system log traces as input, LAR automatically learns and refines system models, and produces two kinds of outputs. One is a counterexample with a bounded probability of being spurious. The other is a probabilistic model based on which the given property is `verified'. The model can be viewed as a proof obligation, i.e., the property is verified if the model is correct. It can also be used for subsequent system analysis activities like runtime monitoring or model-based testing. Our method has been implemented as a self-contained software toolkit. The evaluation on multiple benchmark systems as well as a real-world water treatment system shows promising results.

Pingfan Kong, Yi Li, Xiaohong Chen et al.

Hybrid systems exhibit both continuous and discrete behavior. Analyzing hybrid systems is known to be hard. Inspired by the idea of concolic testing (of programs), we investigate whether we can combine random sampling and symbolic execution in order to effectively verify hybrid systems. We identify a sufficient condition under which such a combination is more effective than random sampling. Furthermore, we analyze different strategies of combining random sampling and symbolic execution and propose an algorithm which allows us to dynamically switch between them so as to reduce the overall cost. Our method has been implemented as a web-based checker named HyChecker. HyChecker has been evaluated with benchmark hybrid systems and a water treatment system in order to test its effectiveness.