Aug 2026· Scientific Reports· Vol 16· 0 citations· 46 references
Medicine
TL;DR
APEW-Fed is presented, a federated ensemble framework that resolves this tension through three provably complementary mechanisms and reconciles heterogeneous score distributions across clients via differentially private quantile sketches under secure aggregation.
Abstract
Enterprise Resource Planning (ERP) systems are high-value targets for fraud owing to the sensitive financial and operational data they process. Anomaly detection in such systems faces a fundamental tension: centralising data for accurate detection conflicts with privacy regulations such as the European Union’s General Data Protection Regulation (GDPR) and the United States’ Sarbanes–Oxley Act (SOX). We present APEW-Fed, a federated ensemble framework that resolves this tension through three provably complementary mechanisms. First, Adaptive Privacy-Aware Ensemble Weighting (APEW) frames model fusion as a privacy-penalised optimisation problem and derives closed-form dynamic weights that up-weight privacy-efficient, high-confidence detectors. Second, Feature-Sensitivity Calibrated Differential Privacy (FS-CDP) partitions features into sensitivity tiers and allocates per-tier Gaussian noise budgets that provably minimise total noise variance subject to a global Rényi DP constraint. Third, Federated Anomaly Score Calibration (FASC) reconciles heterogeneous score distributions across clients via differentially private quantile sketches under secure aggregation. Integrating Isolation Forest, a federated autoencoder trained with per-step DP-SGD, adaptive DBSCAN, and gradient boosting, we evaluate on 243,531 real-world ERP transactions with 5000 expert-annotated anomalies and validate on two public benchmarks: the IEEE-CIS Fraud Detection dataset and the NSL-KDD network intrusion dataset. APEW-Fed achieves 93.5% F1-score at \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$(\alpha ,\epsilon )=(8,1.0)$$\end{document} RDP—only 2.7 percentage points below the non-private centralised baseline—while reducing communication cost by 94% versus FedAvg. On a strictly unbiased uniform-only evaluation subset (2500 transactions, no score-based enrichment), APEW-Fed achieves 92.1% F1, confirming that the headline result is not an artefact of the annotation sampling design. Membership inference attack success is 51.2%, near-random guessing, confirming strong empirical privacy.
Federated learning allows financial institutions to collaboratively identify fraud without distributing raw transaction data, while differential privacy safeguards individual records from inference attacks. Utilizing a lightweight four-layer neural network that was trained on a 10,000-sample subset of the PaySim mobile money dataset (originally 6.3 million transactions), this paper provides a systematic empirical evaluation of a differentially private federated fraud detection framework. The framework is evaluated in the context of varying data distributions (IID and multiple non-IID schemes), clients' numbers, participation rates, privacy budgets, and data quantity skews. The findings indicate that IID and moderately heterogeneous settings produce high accuracy and F1-scores. Conversely, performance is substantially undermined by severe non-IID partitions, numerous small clients, and extreme data imbalances, particularly when differential privacy noise is implemented. The non-private model's detection capability is largely preserved by intermediate privacy budgets (approximately ε≈1). Notably, membership inference attack success rates remain at or near random guessing (approximately 50% accuracy) in the absence of differential privacy, indicating that the baseline provides a restricted attack surface. In this evaluation setting, the privacy benefit is further confirmed by the fact that differential privacy at ε≈1 further suppresses the attack advantage toward zero. While these results provide practical configuration advice for federated, privacy-preserving fraud detection systems, they should be interpreted within the context of the simplified experimental setup that was implemented.
Gian Maxmillian Firdaus, M. Abdurohman, B. Erfianto et al.· International Conference on...· 0 citations
The rapid growth of large data-center (DC) loads is creating new challenges for power-system visibility, privacy, and cyber-physical security. System operators need accurate short-term information about these fast-varying loads, while DC operators may avoid sharing raw megawatt measurements because they can reveal sensitive workload and utilization patterns. This paper presents FedGuard-DC, a federated learning (FL) framework for privacy-preserving DC load forecasting and local false-data-injection attack (FDIA) detection. Each DC trains a dual-head model on its own measurements, where a shared encoder supports both a forecasting head and a reconstruction head. A calibrated anomaly score combines forecast residual and reconstruction error to detect corrupted measurements locally. Raw measurements and absolute MW demand remain at each DC, while only model updates are shared with the global controller. Optional differential privacy and robust trimmed-mean aggregation are included to evaluate privacy-utility behavior and poisoned-client resilience. The framework is validated using EMT simulation data from four large DC loads rated between 150 and 350 MW integrated into the IEEE 39-bus New England system. Results show a 0.5 s-ahead normalized forecast RMSE of 0.023-0.038 pu, compared with 0.32-0.34 pu for persistence. FedGuard-DC detects FDIA with ROC-AUC of 0.979, F1 = 0.930, and precision of 0.988, while robust aggregation reduces the poisoned-client RMSE impact from 0.042 to 0.035 pu.
Context—Card fraud cost the global financial system over $28 billion in 2019, and losses have risen every year since. Banks hold the transaction data needed for collaborative fraud detection, but privacy regulations such as GDPR and KVKK prevent cross-institutional data sharing. Federated Learning keeps raw data local: each institution trains on its own records and sends only model parameters to a shared coordinator. Differentially-Private Stochastic Gradient Descent (DP-SGD) counters gradient inversion attacks by clipping per-sample gradients and adding calibrated noise before parameters leave the client, yielding a record-level (ε, δ) guarantee whose strength depends on the accumulated privacy budget ε. How the aggregation strategy behaves under this noise has not yet been studied.Objective—We compare five aggregation strategies—FedAvg, FedProx, cosine-similarity aggregation, FedAvg-DWA, and FedAdam—across two real-world benchmarks, two heterogeneity levels, four DP-SGD noise multipliers, and 10 random seeds per configuration. We ask whether the relative ranking of strategies survives DP noise, and what DP actually costs in deployed model behavior.Method—510 federated training runs cover five aggregation configurations, two datasets (Kaggle ULB: n = 284807, 0.17% positive rate; IEEE-CIS: n = 590540, 3.50%), two Dirichlet non-IID levels (α ∈ {0.5, 0.1}), and σ ∈ {0.5, 1.0, 1.5, 2.0}. The base learner is a three-layer MLP with LayerNorm. Privacy accounting uses the Opacus RDP accountant at δ = 10⁻⁵. Performance is reported at both the fixed 0.5 threshold and a validation-tuned F₁-maximizing threshold. All pairwise comparisons carry Bonferroni-, Holm-, and Benjamini–Hochberg-corrected p-values, and all 200 realized client-shard partitions are characterized directly.Results—On Kaggle ULB under IID partitioning, federated F₁@best-t (0.801–0.804) matches the centralized baseline (0.799 ± 0.036). Under mild non-IID without DP (α = 0.5), FedAvg-DWA narrowly leads (< 0.015 on F₁@best-t), though neither test survives correction. In the ULB privacy sweep, 7 of 40 Wilcoxon tests show nominal significance but none survive family-wise correction; performance consistently orders with FedAvg-DWA leading and FedAdam trailing. On IEEE-CIS, this ordering amplifies, and the extreme pair remains Holm-significant under DP (adjusted p = 0.020). While AUC holds near 0.95, calibration drift collapses precision at the fixed 0.5 threshold; a validation-tuned threshold recovers F₁ loss without privacy cost. Both calibration effects replicate on IEEE-CIS (ɛ = 8.10 ± 5.30 at σ = 1.0).Conclusion—Aggregation-rule differences under DP-SGD are small and dataset-dependent in magnitude, consistent in direction, and on the ULB sweep none survives correction; outcome differences arise mainly from post-training threshold tuning. We also uncover a previously unreported interaction: Opacus's DP step quietly overrides the standard loss-augmented FedProx setup, making FedProx numerically equivalent to FedAvg, verified by bit-identical per-round global weight trajectories across seeds.
Unknown authors· Pamukkale Üniversitesi Mühen...· 0 citations
Government electronic payment gateways concentrate public revenue and citizen data, making them high-value targets whose compromise carries national-security consequences. Yet vulnerability assessment in many public institutions remains manual, severity-driven, and unable to scale against both the growing volume of disclosed weaknesses and the continuous stream of live attack traffic. To address both vulnerability surfaces within a single coherent assessment layer, this paper presents a dual-module machine learning framework that unifies static prioritisation of disclosed vulnerabilities with runtime detection of malicious traffic. Methodologically, the static module prioritises disclosed vulnerabilities by pairing Term Frequency–Inverse Document Frequency (TF-IDF) text representation with Bayesian-optimised XGBoost multi-class classification over a corpus of 386,337 records, built by expanding National Vulnerability Database (NVD) records across their affected libraries using Open Source Vulnerabilities (OSV) data; the runtime module detects malicious HTTP requests using twelve engineered request features and Bayesian-optimised XGBoost binary classification on the CSIC 2010 dataset of 61,065 requests. On the key results, the static module attains a test accuracy of 0.8630, a macro-averaged F1-score of 0.8475, and High-tier recall of 0.9356; the runtime module attains an accuracy of 0.9101, an F1-score of 0.8879, and a ROC-AUC of 0.9796, raising attack recall from 0.087 under a rule-based signature baseline to 0.868. These results demonstrate that interpretable, low-cost learning models can strengthen vulnerability assessment for government payment infrastructure in resource-constrained settings, contributing the first data-driven, reproducible vulnerability-assessment model developed specifically for the Tanzanian Government Electronic Payment Gateway (GePG) context. As principal limitations, we identify the proxy-label limitation, on which the static module is trained against library-exposure breadth rather than a severity or exploitation measure, and the need for institution-specific validation on live GePG traffic, which together define the principal paths for refinement.
Unknown authors· East African Journal of Info...· 0 citations
Predictive customer behavioral modeling has long assumed centralized access to raw interaction data an assumption that regulatory frameworks, competitive constraints, and cross-organizational data governance requirements render increasingly untenable in contemporary enterprise environments. Federated learning offers a principled alternative, enabling collaborative model training across distributed data holders without centralizing behavioral records. However, standard federated learning frameworks were designed for device-level settings whose structural properties differ substantially from those of customer behavioral data: interaction sequences are longer and sparser, distributions across organizational participants are more heterogeneous, and privacy sensitivities are more legally consequential. The paper proposes a systematic analysis framework for investigating five structural aspects, which need behavioral-specific adjustment beyond regular federated learning approaches in the following contexts: feature engineering with data locality; communication efficiency during distributed behavioral model training; differential privacy in behavioral prediction pipelines; non-IID distribution of behaviors in cross-silo federations; and secure aggregation with Byzantine resilience. For each dimension, the article identifies the specific failure modes that arise when canonical methods are applied without adaptation and provides practitioner-oriented design guidance. The analysis further proposes a deployment-prioritized research agenda whose sequencing is determined by the severity with which unresolved challenges block production deployment. The framework contributes both a diagnostic lens for organizations evaluating federated behavioral intelligence adoption and a structured research roadmap for the methods community.
Kali Prasad Chiruvelli· International Journal of Com...· 0 citations
Federated learning (FL) allows multiple institutions to train a shared model without exchanging raw data, which makes it attractive for privacy-sensitive domains such as finance. Deploying FL across jurisdictions, however, remains difficult: privacy regimes such as the General Data Protection Regulation (GDPR) in Europe and the California Consumer Privacy Act (CCPA) impose different, and sometimes conflicting, obligations on data handling. This paper proposes a regulatory-driven FL framework that treats compliance and fairness as first-class optimization objectives rather than as afterthoughts. The framework combines locally trained models with record-level differential privacy, secure aggregation of masked model updates, and a multi-objective loss that balances predictive accuracy, jurisdiction-specific compliance constraints, and group-fairness criteria, followed by a post-hoc bias audit that corrects residual disparities. We implement the complete pipeline and evaluate it on a synthetic cross-jurisdictional credit-scoring benchmark with three regions governed by GDPR-, CCPA-, and locally-styled privacy budgets, comparing against centralized learning and baseline federated averaging over five random seeds. The proposed framework keeps every region within its differential-privacy budget (ε = 1.58, 2.41, and 3.32 against caps of 2.0, 3.0, and 4.0 at δ = 10⁻⁵), raises the disparate impact ratio from 0.88 to 0.99, reduces the demographic-parity difference from 0.068 to 0.007, and attains a perfect score on an operational compliance rubric, while giving up only 0.9 percentage points of accuracy (88.2% versus 89.1%) at roughly 2.6 times the training cost. The results also surface a known tension: enforcing demographic parity increases the equalized-odds gap, quantifying the price of fairness under differential privacy in federated financial systems.
Jagannath E. Nalavade, R. Sajjan, K. Bhosle et al.· Journal of Intelligent Decis...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
