Nov 2026· International Journal on Advanced Computer Engineering and Communication Technology· Vol 15, pp. 316-324· 0 citations· 34 references
TL;DR
A defense-in-depth framework that coordinates Federated Proximal optimization, record-level differential privacy (DP), Cheon–Kim–Kim–Kim–Song (CKKS) approximate homomorphic encryption, and explainable artificial intelligence (XAI) is proposed.
Abstract
Cross-institutional clinical artificial intelligence is constrained by fragmented patient data, non-identically distributed (non-IID) populations, leakage from shared model updates, and limited interpretability. Federated learning (FL) preserves data locality, but ordinary FL neither makes model updates confidential nor guarantees stable optimization under heterogeneous clients. This paper proposes a defense-in-depth framework that coordinates Federated Proximal (FedProx) optimization, record-level differential privacy (DP), Cheon–Kim–Kim–Song (CKKS) approximate homomorphic encryption, and explainable artificial intelligence (XAI). FedProx limits local model drift; DP-SGD bounds the influence of an individual training record; and CKKS permits weighted aggregation while client updates remain encrypted. A key-separated design keeps the aggregation server from being provisioned with the secret key used to reveal individual contributions; only the aggregate is released for global-model construction. Grad-CAM and SHAP/LIME provide clinician-facing explanations, while an audit dashboard reports convergence, privacy-accounting inputs, encryption status, and systems overhead. We formalize the learning objective, privacy mechanism, encrypted aggregation, threat model, numerical approximation error, communication footprint, and an evaluation protocol spanning IID and controlled non-IID partitions. The contribution is a reproducible system-level co-design and evaluation framework rather than a new cryptographic primitive or optimization algorithm.
This approach holistically rethinks the entire FL-SMPC pipeline, incorporating concurrency at critical stages to improve runtime efficiency while preserving strong security guarantees for the aggregated global model on the server side-an often overlooked aspect in existing SMPC approaches.
Benjamin Hamm, Yannick Kirchhoff, M. Rokuss et al.· IEEE journal of biomedical a...· 2 citations
Aegis is proposed, a principled client-side defense that breaks this dilemma without perturbing patient data or modifying the FL protocol, offering a practical privacy primitive for medical FL.
Chao-Yu Zhang, Shang-Hao Shi, Heng Jin et al.· arXiv.org· 0 citations
Artificial intelligence (AI) in healthcare has had a significant impact on the treatment of diseases, personalized treatment, and clinical decision-making due to its rapid adoption. Nonetheless, the sensitivity of healthcare data poses crucial threats to the privacy of data, data security, and distrust in black-box AI...
Tanishka Rastogi, S. Puri, Anshu Vashisth et al.· International Conference on...· 0 citations
The results demonstrate that strong privacy protection, communication efficiency, and high diagnostic reliability can coexist within large-scale federated healthcare systems, making FedDeepRiskNet + + a practical solution for next-generation medical IoT systems.
Ibrahim Aqeel, Emad S. Hassan, Noha Mostafa et al.· Cluster Computing· 1 citation
BackgroundFederated learning (FL) enables collaborative clinical modeling without centralizing sensitive patient records. However, shared model information may expose private data, while malicious client contributions can compromise model integrity. Combining privacy protection, attack-aware screening, and encrypted ag...
Emad S. Hassan, Noha Mostafa, Ayman A. Alharbi· Technology and Health Care· 0 citations
Federated learning allows for collaborative model training without the need to share raw data, privacy protection, non-IID data, unreliable participants, and limited interpretability are all complex challenges that are intertwined. This study presents an Explainable Federated Intelligence (EFI) framework for trustworth...
Udit Mamodiya, I. Kishor· Journal of Computational Mat...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.