Search

Machine learning (ML) for text classification has been widely used in various domains, such as toxicity detection, chatbot consulting, and review analysis. These applications can significantly impact ethics, economics, and human behavior, raising serious concerns about trusting ML decisions. Several studies indicate that traditional metrics, such as model confidence and accuracy, are insufficient to build human trust in ML models. These models often learn spurious correlations during training and predict based on them during inference. In the real world, where such correlations are absent, their performance can deteriorate significantly. To avoid this, a common practice is to test whether predictions are reasonable. Along with this, a challenge known as the trustworthiness oracle problem has been introduced. Due to the lack of automated trustworthiness oracles, the assessment requires manual validation of the decision process disclosed by explanation methods, which is time-consuming and not scalable. We propose TOKI, the first automated trustworthiness oracle generation method for text classifiers, which automatically checks whether the prediction-contributing words are related to the predicted class using explanation methods and word embeddings. To demonstrate its practical usefulness, we introduce a novel adversarial attack method targeting trustworthiness issues identified by TOKI. We compare TOKI with a naive baseline based solely on model confidence using human-created ground truths of 6,000 predictions. We also compare TOKI-guided adversarial attack method with A2T, a SOTA adversarial attack method. Results show that relying on prediction uncertainty cannot distinguish between trustworthy and untrustworthy predictions, TOKI achieves 142% higher accuracy than the naive baseline, and TOKI-guided adversarial attack method is more effective with fewer perturbations than A2T.

Neural code generation systems have recently attracted increasing attention to improve developer productivity and speed up software development. Typically, these systems maintain a pre-trained neural model and make it available to general users as a service (e.g., through remote APIs) and incorporate a feedback mechanism to extensively collect and utilize the users' reaction to the generated code, i.e., user feedback. However, the security implications of such feedback have not yet been explored. With a systematic study of current feedback mechanisms, we find that feedback makes these systems vulnerable to feedback data injection (FDI) attacks. We discuss the methodology of FDI attacks and present a pre-attack profiling strategy to infer the attack constraints of a targeted system in the black-box setting. We demonstrate two proof-of-concept examples utilizing the FDI attack surface to implement prompt injection attacks and backdoor attacks on practical neural code generation systems. The attacker may stealthily manipulate a neural code generation system to generate code with vulnerabilities, attack payload, and malicious and spam messages. Our findings reveal the security implications of feedback mechanisms in neural code generation systems, paving the way for increasing their security., Comment: Accepted by ISSTA'24

Unanticipated runtime errors, lacking predefined handlers, can abruptly terminate execution and lead to severe consequences, such as data loss or system crashes. Despite extensive efforts to identify potential errors during the development phase, such unanticipated errors remain a challenge to to be entirely eliminated, making the runtime mitigation measurements still indispensable to minimize their impact. Automated self-healing techniques, such as reusing existing handlers, have been investigated to reduce the loss coming through with the execution termination. However, the usability of existing methods is retained by their predefined heuristic rules and they fail to handle diverse runtime errors adaptively. Recently, the advent of Large Language Models (LLMs) has opened new avenues for addressing this problem. Inspired by their remarkable capabilities in understanding and generating code, we propose to deal with the runtime errors in a real-time manner using LLMs. Specifically, we propose Healer, the first LLM-assisted self-healing framework for handling runtime errors. When an unhandled runtime error occurs, Healer will be activated to generate a piece of error-handling code with the help of its internal LLM and the code will be executed inside the runtime environment owned by the framework to obtain a rectified program state from which the program should continue its execution. Our exploratory study evaluates the performance of Healer using four different code benchmarks and three state-of-the-art LLMs, GPT-3.5, GPT-4, and CodeQwen-7B. Results show that, without the need for any fine-tuning, GPT-4 can successfully help programs recover from 72.8% of runtime errors, highlighting the potential of LLMs in handling runtime errors.

In this paper, we test the hypothesis that although OpenAI's GPT-4 performs well generally, we can fine-tune open-source models to outperform GPT-4 in smart contract vulnerability detection. We fine-tune two models from Meta's Code Llama and a dataset of 17k prompts, Detect Llama - Foundation and Detect Llama - Instruct, and we also fine-tune OpenAI's GPT-3.5 Turbo model (GPT-3.5FT). We then evaluate these models, plus a random baseline, on a testset we develop against GPT-4, and GPT-4 Turbo's, detection of eight vulnerabilities from the dataset and the two top identified vulnerabilities - and their weighted F1 scores. We find that for binary classification (i.e., is this smart contract vulnerable?), our two best-performing models, GPT-3.5FT and Detect Llama - Foundation, achieve F1 scores of $0.776$ and $0.68$, outperforming both GPT-4 and GPT-4 Turbo, $0.66$ and $0.675$. For the evaluation against individual vulnerability identification, our top two models, GPT-3.5FT and Detect Llama - Foundation, both significantly outperformed GPT-4 and GPT-4 Turbo in both weighted F1 for all vulnerabilities ($0.61$ and $0.56$ respectively against GPT-4's $0.218$ and GPT-4 Turbo's $0.243$) and weighted F1 for the top two identified vulnerabilities ($0.719$ for GPT-3.5FT, $0.674$ for Detect Llama - Foundation against GPT-4's $0.363$ and GPT-4 Turbo's $0.429$).

Task automation has been greatly empowered by the recent advances in Large Language Models (LLMs) via Python code, where the tasks ranging from software engineering development to general-purpose reasoning. While current benchmarks have shown that LLMs can solve tasks using programs like human developers, the majority of their evaluations are limited to short and self-contained algorithmic tasks or standalone function calls. Solving challenging and practical requires the capability of utilizing diverse function calls as tools to efficiently implement functionalities like data analysis and web development. In addition, using multiple tools to solve a task needs compositional reasoning by accurately understanding complex instructions. Fulfilling both of these characteristics can pose a great challenge for LLMs.To assess how well LLMs can solve challenging and practical tasks via programs, we introduce BigCodeBench, a benchmark that challenges LLMs to invoke multiple function calls as tools from 139 libraries and 7 domains for 1,140 fine-grained tasks. To evaluate LLMs rigorously, each task encompasses 5.6 test cases with an average branch coverage of 99%. In addition, we propose a natural-language-oriented variant of BigCodeBench, BigCodeBench-Instruct, that automatically transforms the original docstrings into short instructions only with essential information. Our extensive evaluation of 60 LLMs shows that LLMs are not yet capable of following complex instructions to use function calls precisely, with scores up to 60%, significantly lower than the human performance of 97%. The results underscore the need for further advancements in this area., Comment: 44 pages, 14 figures, 7 tables, built with love by the BigCode community :)

Artificial Intelligence (AI) models have emerged as another important audience for programming languages alongside humans and machines, as we enter the era of large language models (LLMs). LLMs can now perform well in coding competitions and even write programs like developers to solve various tasks, including mathematical problems. However, the grammar and layout of current programs are designed to cater the needs of human developers -- with many grammar tokens and formatting tokens being used to make the code easier for humans to read. While this is helpful, such a design adds unnecessary computational work for LLMs, as each token they either use or produce consumes computational resources. To improve inference efficiency and reduce computational costs, we propose the concept of AI-oriented grammar. This aims to represent code in a way that better suits the working mechanism of AI models. Code written with AI-oriented grammar discards formats and uses a minimum number of tokens to convey code semantics effectively. To demonstrate the feasibility of this concept, we explore and implement the first AI-oriented grammar for Python, named SimPy. SimPy is crafted by revising the original Python grammar through a series of heuristic rules. Programs written in SimPy maintain identical AST structures to those in standard Python. This allows for not only execution via a modified AST parser, but also seamless transformation between programs written in Python and SimPy, enabling human developers and LLMs to use Python and SimPy, respectively, when they need to collaborate. In the experiments, compared with Python, SimPy enables a reduction in token usage by 13.5% and 10.4% for CodeLlama and GPT-4, respectively, when completing the same set of code-related tasks. Additionally, these models can maintain or even improve their performance when using SimPy instead of Python for these tasks., Comment: Accepted by ISSTA'24

Automated Program Repair (APR) aims to automatically generate patches for rectifying software bugs. Recent strides in Large Language Models (LLM), such as ChatGPT, have yielded encouraging outcomes in APR, especially within the conversation-driven APR framework. Nevertheless, the efficacy of conversation-driven APR is contingent on the quality of the feedback information. In this paper, we propose ContrastRepair, a novel conversation-based APR approach that augments conversation-driven APR by providing LLMs with contrastive test pairs. A test pair consists of a failing test and a passing test, which offer contrastive feedback to the LLM. Our key insight is to minimize the difference between the generated passing test and the given failing test, which can better isolate the root causes of bugs. By providing informative and specific feedback, ContrastRepair enables the LLM to produce effective bug fixes. The implementation of ContrastRepair is based on the state-of-the-art LLM, ChatGPT, and it iteratively interacts with ChatGPT until plausible patches are generated. We evaluate ContrastRepair on multiple benchmark datasets, including Defects4j, QuixBugs, and HumanEval-Java. The results demonstrate that ContrastRepair significantly outperforms existing methods, achieving a new state-of-the-art in program repair. For instance, among Defects4j 1.2 and 2.0, ContrastRepair correctly repairs 143 out of all 337 bug cases, while the best-performing baseline fixes 124 bugs.

Singing voice conversion (SVC) automates song covers by converting a source singing voice from a source singer into a new singing voice with the same lyrics and melody as the source, but sounds like being covered by the target singer of some given target singing voices. However, it raises serious concerns about copyright and civil right infringements. We propose SongBsAb, the first proactive approach to tackle SVC-based illegal song covers. SongBsAb adds perturbations to singing voices before releasing them, so that when they are used, the process of SVC will be interfered, leading to unexpected singing voices. Perturbations are carefully crafted to (1) provide a dual prevention, i.e., preventing the singing voice from being used as the source and target singing voice in SVC, by proposing a gender-transformation loss and a high/low hierarchy multi-target loss, respectively; and (2) be harmless, i.e., no side-effect on the enjoyment of protected songs, by refining a psychoacoustic model-based loss with the backing track as an additional masker, a unique accompanying element for singing voices compared to ordinary speech voices. We also adopt a frame-level interaction reduction-based loss and encoder ensemble to enhance the transferability of SongBsAb to unknown SVC models. We demonstrate the prevention effectiveness, harmlessness, and robustness of SongBsAb on five diverse and promising SVC models, using both English and Chinese datasets, and both objective and human study-based subjective metrics. Our work fosters an emerging research direction for mitigating illegal automated song covers., Comment: In Proceedings of the 32nd Network and Distributed System Security (NDSS) Symposium 2025

Collecting relevant and high-quality data is integral to the development of effective Software Vulnerability (SV) prediction models. Most of the current SV datasets rely on SV-fixing commits to extract vulnerable functions and lines. However, none of these datasets have considered latent SVs existing between the introduction and fix of the collected SVs. There is also little known about the usefulness of these latent SVs for SV prediction. To bridge these gaps, we conduct a large-scale study on the latent vulnerable functions in two commonly used SV datasets and their utilization for function-level and line-level SV predictions. Leveraging the state-of-the-art SZZ algorithm, we identify more than 100k latent vulnerable functions in the studied datasets. We find that these latent functions can increase the number of SVs by 4x on average and correct up to 5k mislabeled functions, yet they have a noise level of around 6%. Despite the noise, we show that the state-of-the-art SV prediction model can significantly benefit from such latent SVs. The improvements are up to 24.5% in the performance (F1-Score) of function-level SV predictions and up to 67% in the effectiveness of localizing vulnerable lines. Overall, our study presents the first promising step toward the use of latent SVs to improve the quality of SV datasets and enhance the performance of SV prediction tasks., Comment: Accepted as a full paper in the technical track at the 21st International Conference on Mining Software Repositories (MSR) 2024

Recent breakthroughs in pre-trained code models, such as CodeBERT and Codex, have shown their superior performance in various downstream tasks. The correctness and unambiguity of API usage among these code models are crucial for achieving desirable program functionalities, requiring them to learn various API fully qualified names structurally and semantically. Recent studies reveal that even state-of-the-art pre-trained code models struggle with suggesting the correct APIs during code generation. However, the reasons for such poor API usage performance are barely investigated. To address this challenge, we propose using knowledge probing as a means of interpreting code models, which uses cloze-style tests to measure the knowledge stored in models. Our comprehensive study examines a code model's capability of understanding API fully qualified names from two different perspectives: API call and API import. Specifically, we reveal that current code models struggle with understanding API names, with pre-training strategies significantly affecting the quality of API name learning. We demonstrate that natural language context can assist code models in locating Python API names and generalize Python API name knowledge to unseen data. Our findings provide insights into the limitations and capabilities of current pre-trained code models, and suggest that incorporating API structure into the pre-training process can improve automated API usage and code representations. This work provides significance for advancing code intelligence practices and direction for future studies. All experiment results, data and source code used in this work are available at \url{https://doi.org/10.5281/zenodo.7902072}.

Code datasets are of immense value for training neural-network-based code completion models, where companies or organizations have made substantial investments to establish and process these datasets. Unluckily, these datasets, either built for proprietary or public usage, face the high risk of unauthorized exploits, resulting from data leakages, license violations, etc. Even worse, the ``black-box'' nature of neural models sets a high barrier for externals to audit their training datasets, which further connives these unauthorized usages. Currently, watermarking methods have been proposed to prohibit inappropriate usage of image and natural language datasets. However, due to domain specificity, they are not directly applicable to code datasets, leaving the copyright protection of this emerging and important field of code data still exposed to threats. To fill this gap, we propose a method, named CodeMark, to embed user-defined imperceptible watermarks into code datasets to trace their usage in training neural code completion models. CodeMark is based on adaptive semantic-preserving transformations, which preserve the exact functionality of the code data and keep the changes covert against rule-breakers. We implement CodeMark in a toolkit and conduct an extensive evaluation of code completion models. CodeMark is validated to fulfill all desired properties of practical watermarks, including harmlessness to model accuracy, verifiability, robustness, and imperceptibility., Comment: Accepted to FSE 2023

The increasingly popular adoption of deep learning models in many critical source code tasks motivates the development of data augmentation (DA) techniques to enhance training data and improve various capabilities (e.g., robustness and generalizability) of these models. Although a series of DA methods have been proposed and tailored for source code models, there lacks a comprehensive survey and examination to understand their effectiveness and implications. This paper fills this gap by conducting a comprehensive and integrative survey of data augmentation for source code, wherein we systematically compile and encapsulate existing literature to provide a comprehensive overview of the field. We start with an introduction of data augmentation in source code and then provide a discussion on major representative approaches. Next, we highlight the general strategies and techniques to optimize the DA quality. Subsequently, we underscore techniques useful in real-world source code scenarios and downstream tasks. Finally, we outline the prevailing challenges and potential opportunities for future research. In essence, we aim to demystify the corpus of existing literature on source code DA for deep learning, and foster further exploration in this sphere. Complementing this, we present a continually updated GitHub repository that hosts a list of update-to-date papers on DA for source code modeling, accessible at \url{https://github.com/terryyz/DataAug4Code}., Comment: ongoing work; 89 publications

Currently, large pre-trained language models are widely applied in neural code completion systems. Though large code models significantly outperform their smaller counterparts, around 70\% of displayed code completions from Github Copilot are not accepted by developers. Being reviewed but not accepted, their help to developer productivity is considerably limited and may conversely aggravate the workload of developers, as the code completions are automatically and actively generated in state-of-the-art code completion systems as developers type out once the service is enabled. Even worse, considering the high cost of the large code models, it is a huge waste of computing resources and energy, which severely goes against the sustainable development principle of AI technologies. However, such waste has never been realized, not to mention effectively addressed, in the research community for neural code completion. Hence, preventing such unhelpful code completions from happening in a cost-friendly way is of urgent need. To fill this significant gap, we first investigate the prompts of unhelpful code completions, called "low-return prompts". We empirically identify four observable patterns in low-return prompts, each lacking necessary information, making it difficult to address through enhancements to the model's accuracy alone. This demonstrates the feasibility of identifying such low-return prompts based on the prompts themselves. Motivated by this finding, we propose an early-rejection mechanism to turn down low-return prompts by foretelling the code completion qualities. The prompts that are estimated to receive unhelpful code completions will not be sent to the model. Furthermore, we investigated five types of estimators to demonstrate the feasibility of the mechanism. The experimental results show that the estimator can reject 20% of code completion requests with a 97.4% Precision., Comment: Accepted by TOSEM

The performance of neural code search is significantly influenced by the quality of the training data from which the neural models are derived. A large corpus of high-quality query and code pairs is demanded to establish a precise mapping from the natural language to the programming language. Due to the limited availability, most widely-used code search datasets are established with compromise, such as using code comments as a replacement of queries. Our empirical study on a famous code search dataset reveals that over one-third of its queries contain noises that make them deviate from natural user queries. Models trained through noisy data are faced with severe performance degradation when applied in real-world scenarios. To improve the dataset quality and make the queries of its samples semantically identical to real user queries is critical for the practical usability of neural code search. In this paper, we propose a data cleaning framework consisting of two subsequent filters: a rule-based syntactic filter and a model-based semantic filter. This is the first framework that applies semantic query cleaning to code search datasets. Experimentally, we evaluated the effectiveness of our framework on two widely-used code search models and three manually-annotated code retrieval benchmarks. Training the popular DeepCS model with the filtered dataset from our framework improves its performance by 19.2% MRR and 21.3% Answer@1, on average with the three validation benchmarks., Comment: 11 pages, accepted to ICSE 2022

Github Copilot, trained on billions of lines of public code, has recently become the buzzword in the computer science research and practice community. Although it is designed to help developers implement safe and effective code with powerful intelligence, practitioners and researchers raise concerns about its ethical and security problems, e.g., should the copyleft licensed code be freely leveraged or insecure code be considered for training in the first place? These problems pose a significant impact on Copilot and other similar products that aim to learn knowledge from large-scale open-source code through deep learning models, which are inevitably on the rise with the fast development of artificial intelligence. To mitigate such impacts, we argue that there is a need to invent effective mechanisms for protecting open-source code from being exploited by deep learning models. Here, we design and implement a prototype, CoProtector, which utilizes data poisoning techniques to arm source code repositories for defending against such exploits. Our large-scale experiments empirically show that CoProtector is effective in achieving its purpose, significantly reducing the performance of Copilot-like deep learning models while being able to stably reveal the secretly embedded watermark backdoors., Comment: 8 pages, accepted to WWW 2022

Over the past decade, deep learning (DL) has been successfully applied to many industrial domain-specific tasks. However, the current state-of-the-art DL software still suffers from quality issues, which raises great concern especially in the context of safety- and security-critical scenarios. Adversarial examples (AEs) represent a typical and important type of defects needed to be urgently addressed, on which a DL software makes incorrect decisions. Such defects occur through either intentional attack or physical-world noise perceived by input sensors, potentially hindering further industry deployment. The intrinsic uncertainty nature of deep learning decisions can be a fundamental reason for its incorrect behavior. Although some testing, adversarial attack and defense techniques have been recently proposed, it still lacks a systematic study to uncover the relationship between AEs and DL uncertainty. In this paper, we conduct a large-scale study towards bridging this gap. We first investigate the capability of multiple uncertainty metrics in differentiating benign examples (BEs) and AEs, which enables to characterize the uncertainty patterns of input data. Then, we identify and categorize the uncertainty patterns of BEs and AEs, and find that while BEs and AEs generated by existing methods do follow common uncertainty patterns, some other uncertainty patterns are largely missed. Based on this, we propose an automated testing technique to generate multiple types of uncommon AEs and BEs that are largely missed by existing techniques. Our further evaluation reveals that the uncommon data generated by our method is hard to be defended by the existing defense techniques with the average defense success rate reduced by 35\%. Our results call for attention and necessity to generate more diverse data for evaluating quality assurance solutions of DL software., Comment: 13 pages

Speaker recognition (SR) is widely used in our daily life as a biometric authentication or identification mechanism. The popularity of SR brings in serious security concerns, as demonstrated by recent adversarial attacks. However, the impacts of such threats in the practical black-box setting are still open, since current attacks consider the white-box setting only. In this paper, we conduct the first comprehensive and systematic study of the adversarial attacks on SR systems (SRSs) to understand their security weakness in the practical blackbox setting. For this purpose, we propose an adversarial attack, named FAKEBOB, to craft adversarial samples. Specifically, we formulate the adversarial sample generation as an optimization problem, incorporated with the confidence of adversarial samples and maximal distortion to balance between the strength and imperceptibility of adversarial voices. One key contribution is to propose a novel algorithm to estimate the score threshold, a feature in SRSs, and use it in the optimization problem to solve the optimization problem. We demonstrate that FAKEBOB achieves 99% targeted attack success rate on both open-source and commercial systems. We further demonstrate that FAKEBOB is also effective on both open-source and commercial systems when playing over the air in the physical world. Moreover, we have conducted a human study which reveals that it is hard for human to differentiate the speakers of the original and adversarial voices. Last but not least, we show that four promising defense methods for adversarial attack from the speech recognition domain become ineffective on SRSs against FAKEBOB, which calls for more effective defense methods. We highlight that our study peeks into the security implications of adversarial attacks on SRSs, and realistically fosters to improve the security robustness of SRSs., Comment: IEEE Symposium on Security and Privacy 2021

Vulnerability identification is crucial to protect the software systems from attacks for cyber security. It is especially important to localize the vulnerable functions among the source code to facilitate the fix. However, it is a challenging and tedious process, and also requires specialized security expertise. Inspired by the work on manually-defined patterns of vulnerabilities from various code representation graphs and the recent advance on graph neural networks, we propose Devign, a general graph neural network based model for graph-level classification through learning on a rich set of code semantic representations. It includes a novel Conv module to efficiently extract useful features in the learned rich node representations for graph-level classification. The model is trained over manually labeled datasets built on 4 diversified large-scale open-source C projects that incorporate high complexity and variety of real source code instead of synthesis code used in previous works. The results of the extensive evaluation on the datasets demonstrate that Devign outperforms the state of the arts significantly with an average of 10.51% higher accuracy and 8.68\% F1 score, increases averagely 4.66% accuracy and 6.37% F1 by the Conv module., Comment: accepted by NeurIPS 2019

Identifying potentially vulnerable locations in a code base is critical as a pre-step for effective vulnerability assessment; i.e., it can greatly help security experts put their time and effort to where it is needed most. Metric-based and pattern-based methods have been presented for identifying vulnerable code. The former relies on machine learning and cannot work well due to the severe imbalance between non-vulnerable and vulnerable code or lack of features to characterize vulnerabilities. The latter needs the prior knowledge of known vulnerabilities and can only identify similar but not new types of vulnerabilities. In this paper, we propose and implement a generic, lightweight and extensible framework, LEOPARD, to identify potentially vulnerable functions through program metrics. LEOPARD requires no prior knowledge about known vulnerabilities. It has two steps by combining two sets of systematically derived metrics. First, it uses complexity metrics to group the functions in a target application into a set of bins. Then, it uses vulnerability metrics to rank the functions in each bin and identifies the top ones as potentially vulnerable. Our experimental results on 11 real-world projects have demonstrated that, LEOPARD can cover 74.0% of vulnerable functions by identifying 20% of functions as vulnerable and outperform machine learning-based and static analysis-based techniques. We further propose three applications of LEOPARD for manual code review and fuzzing, through which we discovered 22 new bugs in real applications like PHP, radare2 and FFmpeg, and eight of them are new vulnerabilities.

Deep learning (DL) defines a data-driven programming paradigm that automatically composes the system decision logic from the training data. In company with the data explosion and hardware acceleration during the past decade, DL achieves tremendous success in many cutting-edge applications. However, even the state-of-the-art DL systems still suffer from quality and reliability issues. It was only until recently that some preliminary progress was made in testing feed-forward DL systems. In contrast to feed-forward DL systems, recurrent neural networks (RNN) follow a very different architectural design, implementing temporal behaviors and memory with loops and internal states. Such stateful nature of RNN contributes to its success in handling sequential inputs such as audio, natural languages and video processing, but also poses new challenges for quality assurance. In this paper, we initiate the very first step towards testing RNN-based stateful DL systems. We model RNN as an abstract state transition system, based on which we define a set of test coverage criteria specialized for stateful DL systems. Moreover, we propose an automated testing framework, DeepCruiser, which systematically generates tests in large scale to uncover defects of stateful DL systems with coverage guidance. Our in-depth evaluation on a state-of-the-art speech-to-text DL system demonstrates the effectiveness of our technique in improving quality and reliability of stateful DL systems.