Decentralized Federated Adversarial Learning for Proactive Anomaly Defense in Consumer-Grade Agricultural Vision Systems

Zhang T., Li Y., Nie J., Fang K., ERCİŞLİ S., Khan M. A.

IEEE Transactions on Consumer Electronics, 2026 (SCI-Expanded, Scopus)

  • Publication Type: Article / Article
  • Publication Date: 2026
  • Doi Number: 10.1109/tce.2026.3706365
  • Journal Name: IEEE Transactions on Consumer Electronics
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Aerospace Database, Compendex, INSPEC, Materials Science & Engineering Collection (ProQuest), Technology Collection (ProQuest)
  • Keywords: Anomaly Detection, Consumer IoT, Decentralized Federated Learning, Proactive Adversarial Defense, Swarm Intelligence
  • Ataturk University Affiliated: Yes

Abstract

Although consumer electronics have enabled the rapid deployment of low cost cotton apical bud recognition systems these traditional centralized terminals face critical privacy and cybersecurity risks notably their vulnerability to adversarial visual deception in open farmlands. To address these issues a novel decentralized framework integrating swarm intelligence with federated adversarial learning named Fed-Cotton-Guard is proposed. First a peer-to-peer swarm architecture is established enabling edge devices to collaboratively train without sharing raw physical images. Second through the introduction of a Min Max game mechanism edge nodes are empowered to proactively generate and defend against adversarial perturbations locally thereby achieving proactive threat hunting. Furthermore a detection strategy based on parameter distribution residuals is designed to proactively intercept malicious parameter poisoning within the decentralized network. Extensive experiments on resource constrained hardware platforms such as the Jetson Nano demonstrate that Fed-Cotton-Guard maintains a high detection accuracy of 96.5% while effectively resisting various adversarial attacks including FGSM and PGD, compared to existing methods this framework significantly enhances collaborative fault tolerance and communication efficiency providing a highly robust.