Cybersecurity vulnerabilities in federated learning

View at publisher Summarise with VIRAT
Authors: Shaik Valli Haseena, Simna Shanavas, N. Brundha, Ayasha
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Federated Learning (FL) has been conceived as a dispersed machine learning paradigm facilitating collaborative learning at edge devices without exposing raw data. The model is amenable to privacy preservation and data protection regulation, for example, General Data Protection Regulation compliance. Yet, more widespread deployment of FL reveals a new and extreme spectrum of cybersecurity risks. These consist of data poisoning attacks that can potentially severely contaminate model integrity, model inversion attacks that can potentially recover sensitive data from exchanged gradients, adversarial manipulations where malicious agents take advantage of model weaknesses, and incidental privacy leakage. The impact and real world implication of these attacks differs, for example, a successful poisoning attack in medicine can result in misdiagnosis, model inversion in the finance sector could leak client confidential data, and adversarial attacks in Internet of Things (IoT) would control autonomous devices with safety consequences. This chapter critically reviews these threats taking into consideration attack feasibility, harm extent, and detectability, inspired by recent case studies illustrating their applicability in real world FL deployments. We also analyze the effectiveness of current state of the art countermeasures like robust aggregation methods, differential privacy, and cryptographic methods like secure multiparty computation and homomorphic encryption. By synthesizing current research on attack paradigms and counterattack architectures, the chapter offers practical knowledge towards constructing secure, robust, and trustworthy FL systems, particularly in high-risk applications like medicine, finance, and critical infrastructure.

Fields of Study & Subjects

Privacy-Preserving Technologies in Data Adversarial Robustness in Machine Learning Cryptography and Data Security

Publication details

Journal / source
CRC Press eBooks
Pages
61–86
Publication date
25 March 2026
Language
EN
Document type
Book chapter