ABSTRACT With the development of Pre‐trained Language Models (PLMs), a wide variety of Chinese Natural Language Processing (NLP) tasks benefit from the Bidirectional Encoder Representations of Transformers (BERT) structure. Unfortunately, Chinese BERT‐based PLMs are vulnerable to adversarial attacks, exposing robustness issues. It inspired numerous defence works devoted to improving the model robustness. However, existing pre‐training methods incur substantial time costs, while fine‐tuning methods need more evidence to demonstrate how well the target model has been defended. This work aims to bridge this gap. We propose RoFuBERT, a robust fine‐tuning framework using fuzz testing for Chinese BERT‐based PLMs. We extract Chinese basis, glyph, pinyin and semantic features of downstream NLP tasks during fine‐tuning. We integrate them into fuzz testing and evaluate the testing completeness of the model under adversarial attacks. Finally, we retrain the model for robust fine‐tuning. Our evaluation shows that, compared with the two baselines, RoFuBERT efficiently improves adversarial robustness of models. RoFuBERT takes only 0.69 times more time, and the reduction in adversarial attack success rate is improved by 37.76%. The modification rate of the adversarial samples that can deceive models is improved by 5.19%, on average.
ABSTRACT With the advancement of wireless technology, smart spaces such as homes and offices are increasingly populated with smart devices, including sensors and actuators. In such environments, user activities generate time‐series data that can be analysed to derive rules for the autonomous activation of actuators such as appliances, lights and switches. Since time‐series data in smart spaces grows continuously, manual annotation becomes impractical, making unsupervised learning techniques a practical choice for supporting automation. However, verifying the outcomes of these unsupervised methods is challenging due to the absence of ground‐truth labels. Moreover, the internal decision‐making processes of most unsupervised algorithms are complex and opaque, making their outputs difficult for end users to validate. These challenges give rise to the ‘oracle problem’, which complicates verification and validation. In this paper, we propose a metamorphic testing approach for verifying a novel sensor grouping technique in smart spaces. The technique employs a Spectral clustering algorithm with graph‐based feature representations derived from time‐series data. Our approach defines 10 metamorphic relations encompassing both verification and validation perspectives. Test cases are generated based on these relations, enabling the sensor grouping technique to be evaluated against variations in input data and clustering parameters. Experimental results demonstrate that the proposed approach effectively identifies implementation flaws in sensor relationship inference methods, assesses the structural consistency of clustering results in the absence of ground‐truth labels and evaluates the robustness of the technique across diverse smart space scenarios.
ABSTRACT Software reliability growth models (SRGMs) are widely used to predict software reliability based on defect discovery data collected during testing or operational phases. However, their predictive accuracy often degrades in data‐scarce environments, such as early‐stage testing or safety‐critical systems. Although cross‐project transfer learning has been explored to mitigate this issue by leveraging data from past projects, its applicability remains limited due to the scarcity and confidentiality of real‐world datasets. To overcome these limitations, we propose deep synthetic cross‐project SRGM (DSC‐SRGM), a novel approach that integrates synthetic data generation with cross‐project transfer learning. Synthetic datasets are generated using widely validated traditional SRGMs to represent defect discovery trends. A cross‐correlation‐based clustering method is applied to identify synthetic datasets with patterns similar to the target project. These datasets are then used to train a deep learning model for reliability prediction. The proposed method is evaluated on 60 real‐world datasets, and its performance is compared with both traditional SRGMs and cross‐project deep learning models trained on real‐world datasets. DSC‐SRGM achieves improvements of up to 98.3% over traditional SRGMs and up to 41.3% over cross‐project deep learning models trained on real‐world datasets across evaluation metrics. However, excessive use of synthetic data or a naive combination of synthetic and real‐world data may degrade prediction performance, highlighting the importance of maintaining an appropriate data balance. These findings indicate that DSC‐SRGM is a promising approach for software reliability prediction in data‐scarce environments.
ABSTRACT Code review is a well‐established practice to ensure code quality, identify potential bugs, promote knowledge sharing and maintain coding standards within a team or organization. This study aims to investigate the specific practices, challenges and information needs encountered when reviewing GUI‐based test artefacts for web applications, which remain poorly understood. We conducted a qualitative interview study with 14 software testing professionals from six different companies to explore the distinct aspects of reviewing GUI‐based test artefacts. We identified four practices, six challenges and four information needs related to reviewing GUI‐based test artefacts. The foremost challenge is the validation of GUI‐based tests under review. Furthermore, challenges concerning levels of abstraction and test robustness were not addressed in related studies. Additionally, participants proposed six potential improvements for tools and practices to better support the code review process. Notably, the absence of standardized practices and the need to run tests locally were common themes across participants. The code review process for GUI‐based test artefacts differs from that of production code, highlighting the need for practices and tools tailored specifically to the unique demands of GUI‐based testing.
ABSTRACT Smartphones have become the most commonly used electronic devices, realizing most of the functions of desktop computers and providing a variety of applications that meet user needs. However, constrained by various factors such as hardware and software, the nature of the mobile environment differs from that of personal computer (PC) and server environments. Traditional models based on the PC environment cannot accurately assess and predict the reliability of mobile applications. This paper proposes a new software reliability growth model (SRGM) that uses the extended Weibull function as the distribution of the total number of faults in the mobile environment. Meanwhile, considering the characteristics of dynamic outbreaks of software faults in the mobile environment, a multiple change‐points (CPs) scheme is proposed to improve the model. Relevant experiments were designed using entropy measures to verify the fitting and prediction performance of the proposed model. In addition, due to the large number of model parameters, a metaheuristic optimization algorithm called the Snake Optimizer is used to estimate the model parameters, and a sensitivity analysis of the model is conducted at the end. Experimental results indicate that the performance of the proposed model is superior to that of other models. The proposed model can assist developers and testers in evaluating the reliability of mobile software.
