Table 22 Summary of privacy-preserving and compliance methods.

From: A federated learning-based privacy-preserving image processing framework for brain tumor detection from CT scans

Compliance requirement

Method employed

Description

Data minimization (GDPR)

Federated learning

Data remains local on edge devices; only model updates are shared.

Anonymization (GDPR/HIPAA)

Differential privacy (DP)

Adds noise to gradients/updates to prevent leakage of individual records.

Data security (HIPAA)

Secure aggregation

Aggregates encrypted updates so server never sees individual updates.

Data integrity (HIPAA)

Blockchain auditing

If integrated, it ensures immutable logging of updates and access trails.

Access control

Role-based & Token Auth

Ensures only authorized agents (clients/servers) can participate.

Right to erasure (GDPR)

Local model forgetting

Individuals can be removed without affecting global model by unpairing.

Transparency & auditability

Model version logging

Logs version changes to ensure full traceability of model decisions.

Back to article page