Synthetic Dataset for Blockchain Federated learning
Abstract
The proposed model intentionally uses a synthetic federated dataset to simulate a decentralized environment, controlled non-IID distributions, adversarial clients, and privacy constraints. These are difficult to obtain from real hospital collaborations due to ethical limitations and strict sharing regulations. A synthetic federated classification dataset used for processing Enhanced Blockchain Federated Learning framework with FedAdam, differential privacy, secure aggregation, attention-based trust weighting, and blockchain verification. The decentralized environment is offered by synthetic dataset where clients possess private, heterogeneous, and non-IID local data. In view of privacy, security, and regulatory constraints, the aspects processed such as IoT nodes, financial branches, or edge device than centralized processing. The dataset describes 12K samples distributed to 20 clients, in which 600 local items per client. The 13 attributes in which 12 are numerical and 1 binary present in each sample where binary label denotes normal and abnormal. The class setup in dataset resembles 60% of majority class and 40% of minority class distributions. A non-IID atmosphere is created to emulate heterogeneity in terms of diverse features, and noise levels. In this, trust resistance is tested on low trust clients who generate noisy or adversarial updates. The raw details are client, allows only model updates to be shared, provides decentralized aggregation, privacy preserving, and…
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.