This study evaluated the adversarial robustness of machine learning-based fraud detection systems by comparing classifier vulnerability profiles and assessing adversarial training as a mitigation strategy. Using the IEEE-CIS Fraud Detection dataset, comprising 590,540 transactions with a fraud incidence of 3.5%, four classifiers—logistic regression, random forest, gradient boosting, and a feed-forward neural network—were trained under identical preprocessing and class-weighting conditions and then subjected to Fast Gradient Sign Method and Projected Gradient Descent attacks at a perturbation budget of 0.02. Adversarial examples were constructed directly using closed-form and backpropagated gradients for the differentiable classifiers and using a logistic regression surrogate for the non-differentiable ensembles, before adversarial training was applied as a post-attack mitigation stage. Logistic regression proved the most adversarially vulnerable architecture, sustaining a 31.12-percentage-point recall loss under Projected Gradient Descent, while adversarial training subsequently restored its recall from 0.39 to 0.999 at an accuracy cost of 0.10 percentage points. Random forest and gradient boosting were not degraded by the surrogate-based attack, indicating that comparative robustness claims for tree-based ensembles require attack methods suited to their non-differentiable structure rather than transfer-based evaluation alone. Within the scope of this single-dataset evaluation, the findings support the adoption of adversarial training for gradient-based fraud detection models and suggest that robustness claims should be accompanied by disclosure of the attack methodology used to establish them.
Ololade Zainab Adesokan, Abiola Omolola Bamsa, O. Obioha-Val et al.· Journal of Engineering Resea...· 0 citations
Artificial intelligence systems increasingly mediate consequential decisions in credit allocation, healthcare triage, employment screening and public administration, yet the data underpinning these decisions is frequently incomplete, mislabelled, stale or quietly altered as it moves through long and opaque pipelines. This review examines governance frameworks intended to preserve the integrity of algorithmic decision data and to align organisational practice with an increasingly dense regulatory landscape spanning the European Union, the United States and international standard-setting bodies. It synthesises literature on data quality theory, documentation artefacts such as datasheets and model cards, blockchain-based provenance mechanisms, algorithmic auditing regimes and sector-specific compliance obligations in finance and healthcare. The review finds that technical solutions for data quality monitoring have matured considerably faster than the institutional arrangements needed to make such monitoring auditable, contestable and legally enforceable, producing a persistent gap between what is technically feasible and what is organisationally practised. It further finds that regulatory instruments, notably the General Data Protection Regulation and the Artificial Intelligence Act, converge on transparency and documentation obligations but diverge on enforcement mechanics, creating compliance friction for organisations operating across jurisdictions. The review proposes a layered governance model integrating data-level controls, documentation practices, human oversight and independent auditing, and identifies future research priorities around interoperable provenance standards, cross-border regulatory harmonisation and the measurement of data integrity as a continuous rather than a point-in-time property.
Pelumi Damola Adeyinka, Emonena Patrick Obrik-Uloho, Olufunke Cynthia Metibemu et al.· Asian Journal of Research in...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
