AI-Based Architecture and Distributed Processing for the Detection and Mitigation of Spoofing Attacks in IoT Networks

Security Spoofing Detection Edge Computing Convolutional Neural Networks.

Authors

  • William Villegas
    william.villegas@udla.edu.ec
    Escuela de Ingeniría en Ciberseguridad, Facultad de Ingenierías y Ciencias Aplicadas, Universidad de Las Américas, Ecuador https://orcid.org/0000-0002-5421-7710
  • Iván Ortiz-Garcés Escuela de Ingeniría en Ciberseguridad, Facultad de Ingenierías y Ciencias Aplicadas, Universidad de Las Américas, Ecuador
  • Jaime Govea Escuela de Ingeniría en Ciberseguridad, Facultad de Ingenierías y Ciencias Aplicadas, Universidad de Las Américas, Ecuador
Vol. 9 No. 3 (2025): June
Research Articles

Downloads

The exponential increase in IoT devices within industrial networks has heightened their exposure to cyberattacks, with spoofing attacks posing one of the most critical threats. These attacks exploit communication vulnerabilities, enabling malicious entities to manipulate network traffic and impersonate legitimate devices, compromising system integrity and security. This study aims to develop an AI-driven detection and mitigation system to enhance IoT network security against spoofing attacks. The proposed approach integrates Convolutional Neural Networks with a distributed processing architecture based on Edge nodes, enabling real-time anomaly detection while reducing computational overhead on central servers. The system was tested in four simulated industrial scenarios involving up to 1,000 IoT devices and multiple concurrent attacks to validate its effectiveness. The evaluation included detection accuracy, response time, and system scalability metrics. Results indicate a detection rate of up to 95% under optimal conditions and 88% in high-density environments. Detection and response times ranged from 150 ms to 220 ms and 300 ms to 450 ms, respectively. Additionally, 97% of compromised devices were successfully isolated, with a false positive rate between 3% and 6%. This study introduces a scalable and adaptive AI-based framework, surpassing traditional machine learning techniques in accuracy, efficiency, and real-time applicability for industrial IoT security.