Open Access

ARTICLE

Detection of False Data Injection Attacks: A Protected Federated Deep Learning Based on Encryption Mechanism

Chenxin Lin¹, Qun Zhou¹, Zhan Wang^2,*, Ximing Fan², Yaochang Xu², Yijia Xu²

1 College of Electrical Engineering, Sichuan University, Chengdu, 610065, China
2 College of Cybersecurity, Sichuan University, Chengdu, 610065, China

* Corresponding Author: Zhan Wang. Email:

Computers, Materials & Continua 2025, 84(3), 5859-5877. https://doi.org/10.32604/cmc.2025.065496

Received 14 March 2025; Accepted 24 June 2025; Issue published 30 July 2025

Abstract

False Data Injection Attack (FDIA), a disruptive cyber threat, is becoming increasingly detrimental to smart grids with the deepening integration of information technology and physical power systems, leading to system unreliability, data integrity loss and operational vulnerability exposure. Given its widespread harm and impact, conducting in-depth research on FDIA detection is vitally important. This paper innovatively introduces a FDIA detection scheme: A Protected Federated Deep Learning (ProFed), which leverages Federated Averaging algorithm (FedAvg) as a foundational framework to fortify data security, harnesses pre-trained enhanced spatial-temporal graph neural networks (STGNN) to perform localized model training and integrates the Cheon-Kim-Kim-Song (CKKS) homomorphic encryption system to secure sensitive information. Simulation tests on IEEE 14-bus and IEEE 118-bus systems demonstrate that our proposed method outperforms other state-of-the-art detection methods across all evaluation metrics, with peak improvements reaching up to 35%.

---

Keywords

---