Deep learning has revolutionized numerous fields, yet the reliability of Deep Neural Networks (DNNs) remains a concern due to their complexity and data dependency. Traditional software fault localization methods, such as Spectrum-based Fault Localization (SBFL), have been adapted for DNNs but often fall short in effectiveness. These methods typically overlook the propagation of faults through neural pathways, resulting in less precise fault detection. Research indicates that examining neural pathways, rather than individual neurons, is crucial because issues in one neuron can affect its entire pathway. By investigating these interconnected pathways, we can better identify and address problems arising from the collective activity of neurons. To address this limitation, we introduce the NP-SBFL method, which leverages Layer-wise Relevance Propagation (LRP) to identify essential faulty neural pathways. Our method explores multiple fault sources to accurately pinpoint faulty neurons by analyzing their interconnections. Additionally, our multi-stage gradient ascent (MGA) technique, an extension of gradient ascent (GA), enables sequential neuron activation to enhance fault detection. We evaluated NP-SBFL-MGA on the well-established MNIST and CIFAR-10 datasets, comparing it to other methods like DeepFault and NP-SBFL-GA, as well as three neuron measures: Tarantula, Ochiai, and Barinel. Our evaluation utilized all training and test samples (60,000 for MNIST and 50,000 for CIFAR-10) and revealed that NP-SBFL-MGA significantly outperformed the baselines in identifying suspicious pathways and generating adversarial inputs. Notably, Tarantula with NP-SBFL-MGA achieved a remarkable 96.75% fault detection rate compared to DeepFault's 89.90%. NP-SBFL-MGA highlights a strong correlation between critical path coverage and the number of failed tests in DNN fault localization.
With the growing interest on Large Language Models (LLMs) for fault localization and program repair, ensuring the integrity and generalizability of the LLM-based methods becomes paramount. The code in existing widely-adopted benchmarks for these tasks was written before the the bloom of LLMs and may be included in the training data of existing popular LLMs, thereby suffering from the threat of data leakage, leading to misleadingly optimistic performance metrics. To address this issue, we introduce "ConDefects", a novel dataset of real faults meticulously curated to eliminate such overlap. ConDefects contains 1,254 Java faulty programs and 1,625 Python faulty programs. All these programs are sourced from the online competition platform AtCoder and were produced between October 2021 and September 2023. We pair each fault with fault locations and the corresponding repaired code versions, making it tailored for in fault localization and program repair related research. We also provide interfaces for selecting subsets based on different time windows and coding task difficulties. While inspired by LLM-based tasks, ConDefects can be adopted for benchmarking ALL types of fault localization and program repair methods. The dataset is publicly available, and a demo video can be found at https://www.youtube.com/watch?v=22j15Hj5ONk.
Fault Localization (FL) aims to automatically localize buggy lines of code, a key first step in many manual and automatic debugging tasks. Previous FL techniques assume the provision of input tests, and often require extensive program analysis, program instrumentation, or data preprocessing. Prior work on deep learning for APR struggles to learn from small datasets and produces limited results on real-world programs. Inspired by the ability of large language models (LLMs) of code to adapt to new tasks based on very few examples, we investigate the applicability of LLMs to line level fault localization. Specifically, we propose to overcome the left-to-right nature of LLMs by fine-tuning a small set of bidirectional adapter layers on top of the representations learned by LLMs to produce LLMAO, the first language model based fault localization approach that locates buggy lines of code without any test coverage information. We fine-tune LLMs with 350 million, 6 billion, and 16 billion parameters on small, manually curated corpora of buggy programs such as the Defects4J corpus. We observe that our technique achieves substantially more confidence in fault localization when built on the larger models, with bug localization performance scaling consistently with the LLM size. Our empirical evaluation shows that LLMAO improves the Top-1 results over the state-of-the-art machine learning fault localization (MLFL) baselines by 2.3%-54.4%, and Top-5 results by 14.4%-35.6%. LLMAO is also the first FL technique trained using a language model architecture that can detect security vulnerabilities down to the code line level.
Informal natural language that describes code functionality, such as code comments or function documentation, may contain substantial information about a programs intent. However, there is typically no guarantee that a programs implementation and natural language documentation are aligned. In the case of a conflict, leveraging information in code-adjacent natural language has the potential to enhance fault localization, debugging, and code trustworthiness. In practice, however, this information is often underutilized due to the inherent ambiguity of natural language which makes natural language intent challenging to check programmatically. The emergent abilities of Large Language Models (LLMs) have the potential to facilitate the translation of natural language intent to programmatically checkable assertions. However, it is unclear if LLMs can correctly translate informal natural language specifications into formal specifications that match programmer intent. Additionally, it is unclear if such translation could be useful in practice. In this paper, we describe nl2postcond, the problem of leveraging LLMs for transforming informal natural language to formal method postconditions, expressed as program assertions. We introduce and validate metrics to measure and compare different nl2postcond approaches, using the correctness and discriminative power of generated postconditions. We then use qualitative and quantitative methods to assess the quality of nl2postcond postconditions, finding that they are generally correct and able to discriminate incorrect code. Finally, we find that nl2postcond via LLMs has the potential to be helpful in practice; nl2postcond generated postconditions were able to catch 64 real-world historical bugs from Defects4J.
Deep neural networks (DNNs) are susceptible to bugs, just like other types of software systems. A significant uptick in using DNN, and its applications in wide-ranging areas, including safety-critical systems, warrant extensive research on software engineering tools for improving the reliability of DNN-based systems. One such tool that has gained significant attention in the recent years is DNN fault localization. This paper revisits mutation-based fault localization in the context of DNN models and proposes a novel technique, named deepmufl, applicable to a wide range of DNN models. We have implemented deepmufl and have evaluated its effectiveness using 109 bugs obtained from StackOverflow. Our results show that deepmufl detects 53/109 of the bugs by ranking the buggy layer in top-1 position, outperforming state-of-the-art static and dynamic DNN fault localization systems that are also designed to target the class of bugs supported by deepmufl. Moreover, we observed that we can halve the fault localization time for a pre-trained model using mutation selection, yet losing only 7.55% of the bugs localized in top-1 position.
Software vulnerabilities can cause numerous problems, including crashes, data loss, and security breaches. These issues greatly compromise quality and can negatively impact the market adoption of software applications and systems. Traditional bug-fixing methods, such as static analysis, often produce false positives. While bounded model checking, a form of Formal Verification (FV), can provide more accurate outcomes compared to static analyzers, it demands substantial resources and significantly hinders developer productivity. Can Machine Learning (ML) achieve accuracy comparable to FV methods and be used in popular instant code completion frameworks in near real-time? In this paper, we introduce SecureFalcon, an innovative model architecture with only 121 million parameters derived from the Falcon-40B model and explicitly tailored for classifying software vulnerabilities. To achieve the best performance, we trained our model using two datasets, namely the FormAI dataset and the FalconVulnDB. The FalconVulnDB is a combination of recent public datasets, namely the SySeVR framework, Draper VDISC, Bigvul, Diversevul, SARD Juliet, and ReVeal datasets. These datasets contain the top 25 most dangerous software weaknesses, such as CWE-119, CWE-120, CWE-476, CWE-122, CWE-190, CWE-121, CWE-78, CWE-787, CWE-20, and CWE-762. SecureFalcon achieves 94% accuracy in binary classification and up to 92% in multiclassification, with instant CPU inference times. It outperforms existing models such as BERT, RoBERTa, CodeBERT, and traditional ML algorithms, promising to push the boundaries of software vulnerability detection and instant code completion frameworks.
We present QNNRepair, the first method in the literature for repairing quantized neural networks (QNNs). QNNRepair aims to improve the accuracy of a neural network model after quantization. It accepts the full-precision and weight-quantized neural networks and a repair dataset of passing and failing tests. At first, QNNRepair applies a software fault localization method to identify the neurons that cause performance degradation during neural network quantization. Then, it formulates the repair problem into a linear programming problem of solving neuron weights parameters, which corrects the QNN's performance on failing tests while not compromising its performance on passing tests. We evaluate QNNRepair with widely used neural network architectures such as MobileNetV2, ResNet, and VGGNet on popular datasets, including high-resolution images. We also compare QNNRepair with the state-of-the-art data-free quantization method SQuant. According to the experiment results, we conclude that QNNRepair is effective in improving the quantized model's performance in most cases. Its repaired models have 24% higher accuracy than SQuant's in the independent validation set, especially for the ImageNet dataset.
As society becomes increasingly reliant on electricity, the reliability requirements for electricity supply continue to rise. In response, transmission/distribution system operators (T/DSOs) must improve their networks and operational practices to reduce the number of interruptions and enhance their fault localization, isolation, and supply restoration processes to minimize fault duration. This paper proposes a machine learning based fault prediction method that aims to predict incipient faults, allowing T/DSOs to take action before the fault occurs and prevent customer outages.
When deploying Deep Neural Networks (DNNs), developers often convert models from one deep learning framework to another (e.g., TensorFlow to PyTorch). However, this process is error-prone and can impact target model accuracy. To identify the extent of such impact, we perform and briefly present a differential analysis against three DNNs widely used for image recognition (MobileNetV2, ResNet101, and InceptionV3) converted across four well-known deep learning frameworks (PyTorch, Keras, TensorFlow (TF), and TFLite), which revealed numerous model crashes and output label discrepancies of up to 100%. To mitigate such errors, we present a novel approach towards fault localization and repair of buggy deep learning framework conversions, focusing on pre-trained image recognition models. Our technique consists of four stages of analysis: 1) conversion tools, 2) model parameters, 3) model hyperparameters, and 4) graph representation. In addition, we propose various strategies towards fault repair of the faults detected. We implement our technique on top of the Apache TVM deep learning compiler, and we test it by conducting a preliminary fault localization analysis for the conversion of InceptionV3 from TF to TFLite. Our approach detected a fault in a common DNN converter tool, which introduced precision errors in weights, reducing model accuracy. After our fault localization, we repaired the issue, reducing our conversion error to zero.
In smart electrical grids, fault detection tasks may have a high impact on society due to their economic and critical implications. In the recent years, numerous smart grid applications, such as defect detection and load forecasting, have embraced data-driven methodologies. The purpose of this study is to investigate the challenges associated with the security of machine learning (ML) applications in the smart grid scenario. Indeed, the robustness and security of these data-driven algorithms have not been extensively studied in relation to all power grid applications. We demonstrate first that the deep neural network method used in the smart grid is susceptible to adversarial perturbation. Then, we highlight how studies on fault localization and type classification illustrate the weaknesses of present ML algorithms in smart grids to various adversarial attacks
Root Cause Analysis (RCA) plays an indispensable role in distributed data system maintenance and operations, as it bridges the gap between fault detection and system recovery. Existing works mainly study multidimensional localization or graph-based root cause localization. This paper opens up the possibilities of exploiting the recently developed framework of explainable AI (XAI) for the purpose of RCA. In particular, we propose BALANCE (BAyesian Linear AttributioN for root CausE localization), which formulates the problem of RCA through the lens of attribution in XAI and seeks to explain the anomalies in the target KPIs by the behavior of the candidate root causes. BALANCE consists of three innovative components. First, we propose a Bayesian multicollinear feature selection (BMFS) model to predict the target KPIs given the candidate root causes in a forward manner while promoting sparsity and concurrently paying attention to the correlation between the candidate root causes. Second, we introduce attribution analysis to compute the attribution score for each candidate in a backward manner. Third, we merge the estimated root causes related to each KPI if there are multiple KPIs. We extensively evaluate the proposed BALANCE method on one synthesis dataset as well as three real-world RCA tasks, that is, bad SQL localization, container fault localization, and fault type diagnosis for Exathlon. Results show that BALANCE outperforms the state-of-the-art (SOTA) methods in terms of accuracy with the least amount of running time, and achieves at least $6\%$ notably higher accuracy than SOTA methods for real tasks. BALANCE has been deployed to production to tackle real-world RCA problems, and the online results further advocate its usage for real-time diagnosis in distributed data systems.
In Federated Learning (FL), clients independently train local models and share them with a central aggregator to build a global model. Impermissibility to access clients' data and collaborative training make FL appealing for applications with data-privacy concerns, such as medical imaging. However, these FL characteristics pose unprecedented challenges for debugging. When a global model's performance deteriorates, identifying the responsible rounds and clients is a major pain point. Developers resort to trial-and-error debugging with subsets of clients, hoping to increase the global model's accuracy or let future FL rounds retune the model, which are time-consuming and costly. We design a systematic fault localization framework, FedDebug, that advances the FL debugging on two novel fronts. First, FedDebug enables interactive debugging of realtime collaborative training in FL by leveraging record and replay techniques to construct a simulation that mirrors live FL. FedDebug's breakpoint can help inspect an FL state (round, client, and global model) and move between rounds and clients' models seamlessly, enabling a fine-grained step-by-step inspection. Second, FedDebug automatically identifies the client(s) responsible for lowering the global model's performance without any testing data and labels--both are essential for existing debugging techniques. FedDebug's strengths come from adapting differential testing in conjunction with neuron activations to determine the client(s) deviating from normal behavior. FedDebug achieves 100% accuracy in finding a single faulty client and 90.3% accuracy in finding multiple faulty clients. FedDebug's interactive debugging incurs 1.2% overhead during training, while it localizes a faulty client in only 2.1% of a round's training time.
The advances in deep learning (DL) techniques have the potential to deliver transformative technological breakthroughs to numerous complex tasks in modern power systems that suffer from increasing uncertainty and nonlinearity. However, the vulnerability of DL has yet to be thoroughly explored in power system tasks under various physical constraints. This work, for the first time, proposes a novel physics-constrained backdoor poisoning attack, which embeds the undetectable attack signal into the learned model and only performs the attack when it encounters the corresponding signal. The paper illustrates the proposed attack on the real-time fault line localization application. Furthermore, the simulation results on the 68-bus power system demonstrate that DL-based fault line localization methods are not robust to our proposed attack, indicating that backdoor poisoning attacks pose real threats to DL implementations in power systems. The proposed attack pipeline can be easily generalized to other power system tasks.
Anomaly detection in large industrial cooling systems is very challenging due to the high data dimensionality, inconsistent sensor recordings, and lack of labels. The state of the art for automated anomaly detection in these systems typically relies on expert knowledge and thresholds. However, data is viewed isolated and complex, multivariate relationships are neglected. In this work, we present an autoencoder based end-to-end workflow for anomaly detection suitable for multivariate time series data in large industrial cooling systems, including explained fault localization and root cause analysis based on expert knowledge. We identify system failures using a threshold on the total reconstruction error (autoencoder reconstruction error including all sensor signals). For fault localization, we compute the individual reconstruction error (autoencoder reconstruction error for each sensor signal) allowing us to identify the signals that contribute most to the total reconstruction error. Expert knowledge is provided via look-up table enabling root-cause analysis and assignment to the affected subsystem. We demonstrated our findings in a cooling system unit including 34 sensors over a 8-months time period using 4-fold cross validation approaches and automatically created labels based on thresholds provided by domain experts. Using 4-fold cross validation, we reached a F1-score of 0.56, whereas the autoencoder results showed a higher consistency score (CS of 0.92) compared to the automatically created labels (CS of 0.62) -- indicating that the anomaly is recognized in a very stable manner. The main anomaly was found by the autoencoder and automatically created labels and was also recorded in the log files. Further, the explained fault localization highlighted the most affected component for the main anomaly in a very consistent manner.
As Deep Learning (DL) systems are widely deployed for mission-critical applications, debugging such systems becomes essential. Most existing works identify and repair suspicious neurons on the trained Deep Neural Network (DNN), which, unfortunately, might be a detour. Specifically, several existing studies have reported that many unsatisfactory behaviors are actually originated from the faults residing in DL programs. Besides, locating faulty neurons is not actionable for developers, while locating the faulty statements in DL programs can provide developers with more useful information for debugging. Though a few recent studies were proposed to pinpoint the faulty statements in DL programs or the training settings (e.g. too large learning rate), they were mainly designed based on predefined rules, leading to many false alarms or false negatives, especially when the faults are beyond their capabilities. In view of these limitations, in this paper, we proposed DeepFD, a learning-based fault diagnosis and localization framework which maps the fault localization task to a learning problem. In particular, it infers the suspicious fault types via monitoring the runtime features extracted during DNN model training and then locates the diagnosed faults in DL programs. It overcomes the limitations by identifying the root causes of faults in DL programs instead of neurons and diagnosing the faults by a learning approach instead of a set of hard-coded rules. The evaluation exhibits the potential of DeepFD. It correctly diagnoses 52% faulty DL programs, compared with around half (27%) achieved by the best state-of-the-art works. Besides, for fault localization, DeepFD also outperforms the existing works, correctly locating 42% faulty programs, which almost doubles the best result (23%) achieved by the existing works.
Neural networks have had discernible achievements in a wide range of applications. The wide-spread adoption also raises the concern of their dependability and reliability. Similar to traditional decision-making programs, neural networks can have defects that need to be repaired. The defects may cause unsafe behaviors, raise security concerns or unjust societal impacts. In this work, we address the problem of repairing a neural network for desirable properties such as fairness and the absence of backdoor. The goal is to construct a neural network that satisfies the property by (minimally) adjusting the given neural network's parameters (i.e., weights). Specifically, we propose CARE (\textbf{CA}usality-based \textbf{RE}pair), a causality-based neural network repair technique that 1) performs causality-based fault localization to identify the `guilty' neurons and 2) optimizes the parameters of the identified neurons to reduce the misbehavior. We have empirically evaluated CARE on various tasks such as backdoor removal, neural network repair for fairness and safety properties. Our experiment results show that CARE is able to repair all neural networks efficiently and effectively. For fairness repair tasks, CARE successfully improves fairness by $61.91\%$ on average. For backdoor removal tasks, CARE reduces the attack success rate from over $98\%$ to less than $1\%$. For safety property repair tasks, CARE reduces the property violation rate to less than $1\%$. Results also show that thanks to the causality-based fault localization, CARE's repair focuses on the misbehavior and preserves the accuracy of the neural networks.
This paper introduces an algorithm able to detect and localize the occurrance of a fault in an Active Distribution Network, using the measurements collected by Phasor Measurement Units (PMUs). First, a basic algorithm that works under the assumption that all grid buses are equipped with a PMU is designed. Then, formal observability conditions that allow detection and localization with a reduced number of PMUs are provided. Based on these conditions, the algorithm is extended to perform correctly when not all network buses are monitored. Moreover, an Optimal Positioning Algorithm, always based on the observability conditions, is designed. This algorithm allows the user to customize the fault localization resolution. The approach is validated through simulations carried out on a benchmark active distribution network.
Deep Neural Networks (DNNs) are used in a wide variety of applications. However, as in any software application, DNN-based apps are afflicted with bugs. Previous work observed that DNN bug fix patterns are different from traditional bug fix patterns. Furthermore, those buggy models are non-trivial to diagnose and fix due to inexplicit errors with several options to fix them. To support developers in locating and fixing bugs, we propose DeepDiagnosis, a novel debugging approach that localizes the faults, reports error symptoms and suggests fixes for DNN programs. In the first phase, our technique monitors a training model, periodically checking for eight types of error conditions. Then, in case of problems, it reports messages containing sufficient information to perform actionable repairs to the model. In the evaluation, we thoroughly examine 444 models -53 real-world from GitHub and Stack Overflow, and 391 curated by AUTOTRAINER. DeepDiagnosis provides superior accuracy when compared to UMLUAT and DeepLocalize. Our technique is faster than AUTOTRAINER for fault localization. The results show that our approach can support additional types of models, while state-of-the-art was only able to handle classification ones. Our technique was able to report bugs that do not manifest as numerical errors during training. Also, it can provide actionable insights for fix whereas DeepLocalize can only report faults that lead to numerical errors during training. DeepDiagnosis manifests the best capabilities of fault detection, bug localization, and symptoms identification when compared to other approaches.
Automatic Speech Recognition (ASR) systems have become ubiquitous. They can be found in a variety of form factors and are increasingly important in our daily lives. As such, ensuring that these systems are equitable to different subgroups of the population is crucial. In this paper, we introduce, AequeVox, an automated testing framework for evaluating the fairness of ASR systems. AequeVox simulates different environments to assess the effectiveness of ASR systems for different populations. In addition, we investigate whether the chosen simulations are comprehensible to humans. We further propose a fault localization technique capable of identifying words that are not robust to these varying environments. Both components of AequeVox are able to operate in the absence of ground truth data. We evaluated AequeVox on speech from four different datasets using three different commercial ASRs. Our experiments reveal that non-native English, female and Nigerian English speakers generate 109%, 528.5% and 156.9% more errors, on average than native English, male and UK Midlands speakers, respectively. Our user study also reveals that 82.9% of the simulations (employed through speech transformations) had a comprehensibility rating above seven (out of ten), with the lowest rating being 6.78. This further validates the fairness violations discovered by AequeVox. Finally, we show that the non-robust words, as predicted by the fault localization technique embodied in AequeVox, show 223.8% more errors than the predicted robust words across all ASRs.
Hitherto statistical type inference systems rely thoroughly on supervised learning approaches, which require laborious manual effort to collect and label large amounts of data. Most Turing-complete imperative languages share similar control- and data-flow structures, which make it possible to transfer knowledge learned from one language to another. In this paper, we propose a cross-lingual transfer learning framework, PLATO, for statistical type inference, which allows us to leverage prior knowledge learned from the labeled dataset of one language and transfer it to the others, e.g., Python to JavaScript, Java to JavaScript, etc. PLATO is powered by a novel kernelized attention mechanism to constrain the attention scope of the backbone Transformer model such that the model is forced to base its prediction on commonly shared features among languages. In addition, we propose the syntax enhancement that augments the learning on the feature overlap among language domains. Furthermore, PLATO can also be used to improve the performance of the conventional supervised learning-based type inference by introducing cross-lingual augmentation, which enables the model to learn more general features across multiple languages. We evaluated PLATO under two settings: 1) under the cross-domain scenario that the target language data is not labeled or labeled partially, the results show that PLATO outperforms the state-of-the-art domain transfer techniques by a large margin, e.g., it improves the Python to TypeScript baseline by +5.40%@EM, +5.40%@weighted-F1, and 2) under the conventional monolingual supervised learning based scenario, PLATO improves the Python baseline by +4.40%@EM, +3.20%@EM (parametric).