Information and Software Technology最新文献

{"title":"MT-Boost: A metamorphic-testing based training method for enhancing the robustness of deep neural network classifiers","authors":"Kun Qiu ,&nbsp;Yu Zhou ,&nbsp;Pak-Lok Poon ,&nbsp;Tsong Yueh Chen","doi":"10.1016/j.infsof.2025.107902","DOIUrl":"10.1016/j.infsof.2025.107902","url":null,"abstract":"<div><h3>Context:</h3><div>In metamorphic testing (MT), a set of metamorphic relations (MRs) are identified to verify whether or not a trained deep neural network (DNN) can produce consistent performance when specific transformations are applied to its input. Most DNNs trained with existing methods often perform poorly with respect to MRs, thereby indicating that these DNNs are not robust.</div></div><div><h3>Objective:</h3><div>To improve DNN’s performance in the context of MT, a set of defined MRs is used to generate training inputs to retrain a DNN model. Our main objective is to develop a method to balance a DNN’s accuracy and robustness with less time consumption and having the capability to cater to multiple MRs.</div></div><div><h3>Methods:</h3><div>In this paper, we introduce our regularization-based method (known as MT-Boost), which uses reinforcement learning to search for the best way of using MRs to generate inputs and express them as loss function regularizers. When developing MT-Boost, we transform the robustness-improving problem into a reinforcement-learning agent’s training problem.</div></div><div><h3>Results:</h3><div>MT-Boost is evaluated on eight DNN models with four popular datasets. MT-Boost achieves the largest robustness improvement for each model and maintains relatively high accuracy performance when compared with seven other baseline methods. Our sensitivity analysis also shows the high stability performance of MT-Boost across four reinforcement-learning algorithms and other hyperparameters.</div></div><div><h3>Conclusion:</h3><div>Experimental results show that MT-Boost is effective and efficient for improving DNN’s robustness.</div></div>","PeriodicalId":54983,"journal":{"name":"Information and Software Technology","volume":"189 ","pages":"Article 107902"},"PeriodicalIF":4.3,"publicationDate":"2025-10-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145322331","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A method for real-world privacy-preserving Android malware detection through Federated Machine Learning","authors":"Giovanni Ciaramella ,&nbsp;Fabio Martinelli ,&nbsp;Christian Peluso ,&nbsp;Antonella Santone ,&nbsp;Francesco Mercaldo","doi":"10.1016/j.infsof.2025.107892","DOIUrl":"10.1016/j.infsof.2025.107892","url":null,"abstract":"<div><div>Privacy is one of the most critical issues associated with spreading the Internet of Things and Internet of Everything devices. Over the years, several methods have been introduced to address this phenomenon. In 2017, Google introduced the concept of Federated Machine Learning. This paradigm allows models to be trained collaboratively across multiple decentralized devices or servers, holding local data samples without exchanging them. This approach enhances data privacy and security by ensuring raw data remains on local devices while only model updates are shared and aggregated. This paper presents a privacy-preserving Android malware detector based on Federated Machine Learning. As a first step, we built a dataset comprising over 40,000 Android applications, including trusted and malicious (belonging to 71 malware families) samples. Afterward, we conducted experiments leveraging three different architectures by exploiting the CIFAR-10 and the ImageNet datasets, employing hyperparameters determined through a Grid Search algorithm by exploiting 40 clients. Moreover, the experimental analysis uses two distributions: Independent and identically distributed and non-independent and identically distributed data. To conclude the Federated Machine Learning experiments, we trained models for each architecture, with both weight types and distribution models, by applying the Clipping Norm Aggregator. The results exhibit interesting performances with Independent and identically distributed data, achieving an accuracy of 0.873 without normalization and 0.877 with the Clipping Norm aggregator. However, with non-independent and identically distributed data, the model accuracy equals 0.865 without normalization, 0.864 with the Clipping Norm aggregator using Custom MobileNet 2. In conclusion, to compare Federated Machine Learning with a centralized training approach, we trained several models adopting the same dataset, dataset splitting, and architectures, achieving an accuracy of 0.944 using InceptionV3. The outcomes show that the proposed method can provide engaging performances in privacy-preserving Android malware detection.</div></div>","PeriodicalId":54983,"journal":{"name":"Information and Software Technology","volume":"189 ","pages":"Article 107892"},"PeriodicalIF":4.3,"publicationDate":"2025-10-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145270840","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Gender-based cognitive bias and design thinking in the work of Finnish IT professionals","authors":"Aila Kronqvist ,&nbsp;Rebekah Rousi","doi":"10.1016/j.infsof.2025.107910","DOIUrl":"10.1016/j.infsof.2025.107910","url":null,"abstract":"<div><h3>Context</h3><div>Cognitive bias is a concern in artificial intelligence (AI) development. Research shows the prominence of cognitive bias within algorithms. We argue that cognitive bias is more than training data, but rather development team composition. Design Thinking (DT) is an approach used to reduce bias via multidisciplinary expertise. The article presents a study examining DT in addressing gender-based cognitive bias in the Finnish information technology industry.</div></div><div><h3>Objective</h3><div>The aim was to examine how the gender of IT professionals influences familiarity with and use of DT, coupled with awareness and addressing of cognitive bias in IT development processes.</div></div><div><h3>Method</h3><div>A mixed method questionnaire was used to collect data from <em>N</em> = 93 participants. Questions probed familiarity with DT, use of DT, and cognitive bias handling in participants’ organizations. Non-parametric tests were used to analyze quantitative data, due to abnormal distributions. Atlas.ti was used to code and analyze the qualitative data. Categorization determined whether participants recognized bias in their work, and the importance they attributed towards dealing with gender-based bias in IT.</div></div><div><h3>Results</h3><div>Women were more likely to view gender-based cognitive bias as relevant. Women were significantly more familiar with DT as a methodology (<em>p</em> = 0.028), men were significantly more likely to engage in user studies (<em>p</em> = 0.018). Older participants showed a tendency to emphasize the importance of open discussion more than other participant groups, with some analyses indicating a trend-level difference (<em>p</em> = 0.085). Qualitative responses indicated the importance of discussion in development teams to avoid or mitigate bias, suggesting the need for organizational psychological safety.</div></div><div><h3>Conclusion</h3><div>The paper provides novel contributions to the human dimension of bias in AI and IT in general. Results show that men and women IT professionals were aware of DT, yet men professionals were more likely to mitigate bias through collecting insight from end-users.</div></div>","PeriodicalId":54983,"journal":{"name":"Information and Software Technology","volume":"189 ","pages":"Article 107910"},"PeriodicalIF":4.3,"publicationDate":"2025-10-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145271472","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"FairST: A novel approach for machine learning bias repair through latent sensitive attribute translation","authors":"Carmen Meinson ,&nbsp;Max Hort ,&nbsp;Federica Sarro","doi":"10.1016/j.infsof.2025.107900","DOIUrl":"10.1016/j.infsof.2025.107900","url":null,"abstract":"<div><h3>Context:</h3><div>As Machine Learning (ML) models are increasingly used in critical decision-making software, concerns have been raised about these systems perpetuating or exacerbating existing historical biases. Consequently, there has been a growing research interest in developing methods to test for fairness and repair biases in ML software, particularly for legally protected attributes like gender, age, race.</div></div><div><h3>Objectives:</h3><div>In this work, we set out to repair bias for both single and multiple protected attributes (a.k.a. intersectional fairness) of pre-trained machine learning models.</div></div><div><h3>Methods:</h3><div>We propose a novel model- and task-agnostic debiasing method, Fair Subgroup Translation (FairST), based on fair representation learning via auto-encoders. To the best of our knowledge, this is the first approach based on the principle of Fair Representation Learning devised for post-processing bias repair.</div></div><div><h3>Results:</h3><div>We empirically evaluate the effectiveness of using FairST to repair a pre-trained Neural Network model used for seven classification tasks involving both single and multiple protected attributes, and benchmark its performance with state-of-the-art fairness repair methods (i.e., Learning Fair Representations, Reweighing, FairBalance and FairMask). We also investigate if the effectiveness of FairST varies when using it to repair bias of other popular ML models (namely Logistic Regression, Support Vector Machine, Gaussian Naive Bayes, Decision Tree and Random Forest).</div></div><div><h3>Conclusion:</h3><div>The results demonstrate that FairST consistently achieves superior single and intersectional fairness with respect to all benchmarking methods for all classification tasks considered in our empirical study. This supports the potential of using FairST for ML bias repair, and opens up a rich agenda of future work including its application to repair bias arising in tasks of a different nature such as multi-class or image-based problems.</div></div>","PeriodicalId":54983,"journal":{"name":"Information and Software Technology","volume":"189 ","pages":"Article 107900"},"PeriodicalIF":4.3,"publicationDate":"2025-10-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145271474","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Edge AI assurance: A systematic mapping study","authors":"Clara Ayora,&nbsp;Arturo S. García,&nbsp;Jose Luis de la Vara","doi":"10.1016/j.infsof.2025.107908","DOIUrl":"10.1016/j.infsof.2025.107908","url":null,"abstract":"<div><h3>Context</h3><div>In critical domains, assurance corresponds to the set of activities to provide confidence that a system can be deemed dependable, e.g., safe and secure. This essential system and software engineering process is usually conducted according to standards. For novel applications running at the edge and using artificial intelligence (AI), how to conduct assurance in a systematic way is still under study.</div></div><div><h3>Objective</h3><div>The goal of this paper is to provide a comprehensive understanding of current Edge AI assurance considerations. Our interest lies in contributing insights that offer a forward-looking perspective on what is essential in this research field.</div></div><div><h3>Method</h3><div>We conducted a systematic mapping study (SMS) to characterize how Edge AI assurance is addressed in existing literature. The SMS was built on 38 studies, selected through a multi-stage process, from 3113 studies published between 2019 and 2025. The 38 studies were deeply analysed focusing on seven research questions about the main key Edge AI assurance aspects: dependability concerns, application domains, assurance standards, assurance evidence, dependability justification techniques, and edge and AI characteristics.</div></div><div><h3>Results</h3><div>We found ten dependability concerns that have been addressed (e.g., safety and security), six application domains (e.g., Industry 4.0), eight assurance standards and regulations (e.g., ISO 26262), 27 types of assurance evidence (e.g., architecture specification), three dependability justification techniques (e.g., argumentation), five AI-specific characteristics (e.g., machine learning algorithms) and five edge-specific characteristics (e.g., network).</div></div><div><h3>Conclusions</h3><div>The paper is, to our knowledge, the only existing review on the topic of Edge AI assurance. The results are relevant to practitioners seeking a better grasp on this field as well as researchers to find new research gaps. We have also identified research areas where more effort can be undertaken (e.g., multi-concern assurance).</div></div>","PeriodicalId":54983,"journal":{"name":"Information and Software Technology","volume":"189 ","pages":"Article 107908"},"PeriodicalIF":4.3,"publicationDate":"2025-10-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145270839","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A multi-criteria accessibility index for quantifying digital fairness in e-commerce","authors":"Adam Wasilewski ,&nbsp;Katarzyna Walecka-Jankowska ,&nbsp;Anna Zgrzywa-Ziemak","doi":"10.1016/j.infsof.2025.107907","DOIUrl":"10.1016/j.infsof.2025.107907","url":null,"abstract":"<div><h3>Context:</h3><div>The research is situated within the debate on fair, sustainable, and inclusive information systems. It focuses on the analysis of the digital accessibility of Amazon’s marketplaces and widely accepted standards, alongside legal differences related to the European Accessibility Act (EAA).</div></div><div><h3>Objective:</h3><div>The study aims to analyze the intersection of digital accessibility and fairness on global e-commerce platforms. In the context of sustainable and inclusive software systems, the study critically assesses the extent to which Amazon adheres to the Web Content Accessibility Guidelines (WCAG) in various regional domains. The purpose of the article is to present the proposed approach to studying digital accessibility for more sustainable and fair e-commerce systems.</div></div><div><h3>Methods:</h3><div>For the purpose of the study, a measure of digital accessibility was proposed, and a quantitative analysis of the level of WCAG compliance was conducted based on the results obtained after analyzing the homepages and selected product pages. The empirical research was repeated twice, six months apart, to observe the direction of change in the context of the need for some marketplaces to comply with EAA regulations.</div></div><div><h3>Results:</h3><div>The analysis reveals significant disparities in accessibility implementation, often correlating with the presence or absence of international regulatory mandates, such as the EAA. These findings highlight a systemic trade-off in which compliance-driven enhancements in certain locations coincide with accessibility deficits elsewhere. The study contributes to the discourse on fairness in software systems by demonstrating how uneven regulatory pressures can result in inequitable user experiences.</div></div><div><h3>Conclusion:</h3><div>Using the proposed aggregate measure, an analysis of changes in the digital accessibility of all Amazon marketplaces over time was conducted, showing that regulations affect the fairness of information systems. In addition, the effectiveness of the introduced metric was verified, opening up possibilities for its wider use to improve the quality of information systems.</div></div>","PeriodicalId":54983,"journal":{"name":"Information and Software Technology","volume":"188 ","pages":"Article 107907"},"PeriodicalIF":4.3,"publicationDate":"2025-09-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145227372","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Towards early detection of algorithmic bias from dataset’s bias symptoms: An empirical study","authors":"Giordano d’Aloisio,&nbsp;Claudio Di Sipio,&nbsp;Antinisca Di Marco,&nbsp;Davide Di Ruscio","doi":"10.1016/j.infsof.2025.107905","DOIUrl":"10.1016/j.infsof.2025.107905","url":null,"abstract":"<div><h3>Context:</h3><div>The rise of AI software has made fairness auditing essential, particularly where biased decisions have serious impacts. This entails identifying sensitive variables and calculating fairness metrics based on predictions from a baseline model. Since model training is computationally intensive, recent research focuses on early bias assessment to detect bias before extensive training starts.</div></div><div><h3>Objective:</h3><div>This paper presents an empirical study to evaluate how dataset statistics, named <em>bias symptoms</em>, can assist in the early identification of variables that may lead to bias in the system. The aim of this study is to avoid training a machine learning model before assessing – and, in case, mitigating – its bias, thus increasing the sustainability of the development process.</div></div><div><h3>Method:</h3><div>We first identify a <em>bias symptoms</em> dataset, employing 24 datasets from diverse application domains commonly used in fairness auditing. Through extensive empirical analysis, we investigate the ability of these <em>bias symptoms</em> to predict variables associated with bias under three fairness definitions.</div></div><div><h3>Results:</h3><div>Our results demonstrate that <em>bias symptoms</em> are effective in supporting early predictions of bias-inducing variables under specific fairness definitions.</div></div><div><h3>Conclusion:</h3><div>These findings offer valuable insights for practitioners and researchers, encouraging further exploration in developing methods for proactive bias mitigation involving bias symptoms.</div></div>","PeriodicalId":54983,"journal":{"name":"Information and Software Technology","volume":"188 ","pages":"Article 107905"},"PeriodicalIF":4.3,"publicationDate":"2025-09-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145227211","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Semantic-aware testing for object detection systems","authors":"Xiaoxia Liu ,&nbsp;Jingyi Wang ,&nbsp;Hsiao-Ying Lin ,&nbsp;Chengfang Fang ,&nbsp;Jie Shi ,&nbsp;Xiaodong Zhang ,&nbsp;Wenhai Wang","doi":"10.1016/j.infsof.2025.107888","DOIUrl":"10.1016/j.infsof.2025.107888","url":null,"abstract":"<div><h3>Context:</h3><div>Deep Learning-based object detection (OD) module is rapidly being the common basis for many popular autonomous systems such as self-driving cars and drones. Comprehensive robust testing is essential before deploying an OD module in safety-critical applications.</div></div><div><h3>Objective:</h3><div>We aim to address the following limitations of existing OD testing works: (1) they focus primarily on 2D OD with single-camera inputs rather than 3D OD with multi-camera fusion; (2) they rely on limited environmental changes or GAN transformations that inadequately cover diverse and complex real-world input; and (3) existing testing metrics remain unevaluated in the OD setting.</div></div><div><h3>Methods:</h3><div>We propose and develop a systematic semantic-aware testing framework named <span>SeaT-OD</span> capable of testing practical 3D OD systems based on fused image input by tackling several key technical challenges. Our approach introduces: (1) novel semantic-aware metrics defined over deep feature spaces applicable across diverse OD models; (2) a test generation algorithm using <em>deep semantic transformation</em> to enhance input semantic coverage; and (3) metric-guided test case selection for efficient model robustness improvement through targeted retraining.</div></div><div><h3>Result:</h3><div>We evaluated <span>SeaT-OD</span> on state-of-the-art commonly adopted 2D and 3D OD models based on fused image input in popular datasets from autonomous driving. Extensive experimental results show that existing OD testing works are insufficient, and <span>SeaT-OD</span> is effective in measuring the adequacy of testing practical 3D OD systems, generating high-quality test cases, and selecting test cases meaningful for improving the system robustness.</div></div><div><h3>Conclusion:</h3><div>Based on the results, we emphasize the importance of testing OD systems. Additionally, we present several observations that can direct future research and developments.</div></div>","PeriodicalId":54983,"journal":{"name":"Information and Software Technology","volume":"189 ","pages":"Article 107888"},"PeriodicalIF":4.3,"publicationDate":"2025-09-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145229865","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"An approach to carve sustainability into architecture knowledge practice","authors":"Markus Funke ,&nbsp;Patricia Lago ,&nbsp;Roel Donker","doi":"10.1016/j.infsof.2025.107904","DOIUrl":"10.1016/j.infsof.2025.107904","url":null,"abstract":"<div><h3>Context:</h3><div>The daily work of software architects relies not only on technical expertise, but also on architecture knowledge (AK), which entails the architecture design itself together with the taken decisions and assumptions that provide the rationale for the chosen software solution. However, there is a pressing need to also incorporate sustainability aspects into the complete AK process, i.e., capturing decisions and the design towards software systems that are environmentally, economically, and socially balanced in the long term.</div></div><div><h3>Objective:</h3><div>The goal of this study is twofold. First, we aim to provide the fundamental concepts currently used in industry practices to represent and communicate AK. Second, building on this understanding and uncovered insights, we want to explore where sustainability can be applied in such daily practice and how sustainability can be addressed in the existing AK process.</div></div><div><h3>Method:</h3><div>We present an empirical study, using the same qualitative research methods applied in two independent industrial contexts. In total, the study encompasses a questionnaire survey with 49 participants, semi-structured interviews with 27 practitioners, and two focus groups.</div></div><div><h3>Results:</h3><div>Based on the insights gained from combining our findings, we (i) provide a map of applied concepts for communicating and representing AK in two large enterprises, (ii) present a set of tangible recommendations, and (iii) introduce an approach consisting of an action plan for carving sustainability into current software architecture practice and a reusable methodology applicable to other domains and companies.</div></div><div><h3>Conclusion:</h3><div>This paper extends our previous work by adding empirical investigations, and by refining as well as extending the results; this leads to a more comprehensive discussion of the implications for the future of software architecture practice.</div></div>","PeriodicalId":54983,"journal":{"name":"Information and Software Technology","volume":"188 ","pages":"Article 107904"},"PeriodicalIF":4.3,"publicationDate":"2025-09-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145227212","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"FAIR-CARE: A comparative evaluation of unfairness mitigation approaches","authors":"Chiara Criscuolo ,&nbsp;Mattia Salnitri ,&nbsp;Davide Martinenghi","doi":"10.1016/j.infsof.2025.107898","DOIUrl":"10.1016/j.infsof.2025.107898","url":null,"abstract":"<div><div>Bias and unfairness in Machine Learning (ML) are challenging to detect and mitigate, particularly in critical fields such as finance, hiring, and healthcare. While numerous unfairness mitigation techniques exist, most evaluation frameworks assess only a limited set of fairness metrics, primarily focusing on the trade-off between fairness and accuracy. We introduce FAIR-CARE, a new open-source and robust approach that consists of an evaluation pipeline designed for the systematic assessment of unfairness mitigation techniques. Our approach simultaneously evaluates multiple fairness and performance metrics across various ML models. We conduct a comparative analysis on healthcare datasets with diverse distributions—including target class, protected attribute, and their joint distributions—to identify the most effective mitigation technique for each processing type (pre-, in-, and post-processing). Furthermore, we determine the best-performing techniques across different datasets, fairness metrics, performance metrics, and ML models. Finally, we provide practical insights into the application of these techniques, offering actionable guidance for both researchers and practitioners.</div></div>","PeriodicalId":54983,"journal":{"name":"Information and Software Technology","volume":"189 ","pages":"Article 107898"},"PeriodicalIF":4.3,"publicationDate":"2025-09-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145229866","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}