HAPV-FEL Transaction Risk Detection Data for IBM AML and SAML-D
Abstract
This dataset package is associated with the study “HAPV-FEL: Historical Consensus Verification and Recoverable Trust for Privacy-Preserving Federated Transaction Risk Detection.” It contains structured CSV representations corresponding to the IBM AML and SAML-D datasets used for cross-institutional transaction risk detection experiments. The data organization follows the experimental setting of 18 federated clients and 120 global training rounds. Each record includes client and round identifiers, transaction-related attributes such as time, amount, currency and payment type, graph-related account features, binary transaction-risk labels, and fields describing the adversarial federated-learning setting. The experimental protocol considers four attack types: label flipping, sign flipping, scaling, and Gaussian/model poisoning, with a nominal malicious-client ratio of 20%. The files are intended to support data inspection, machine-learning experiments, federated transaction-risk analysis, and reproduction of the experimental structure described in the study. Because the manuscript does not publish the complete record-level raw IBM AML and SAML-D data used in the experiments, the CSV records provided here are structured reconstructions based on the reported experimental configuration and should not be interpreted as the original raw transaction records.
Read the paper
Related studies
- Active Learning during Federated Learning for Object Detection
- Fairness-aware federated learning
- Decentralized Federated Learning
- Building a Personally Identifiable Information Recognizer in a Privacy Preserved Manner Using Automated Annotation and Federated Learning
- Preservation of Patient Level Privacy: Federated Classification and Calibration Models
- Efficient explainable federated learning
Go deeper in Evidence Hub
Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.