ENHANCING IOT SECURITY THROUGH ML-BASED ADVERSARIAL ATTACK DETECTION

Abstract

Adversarial attacks are an increasing concern for AI systems, especially in IoT, where devices rely on accurate and consistent data for decision-making. This project proposes a lightweight machine learning approach to detect such attacks, using reliable algorithms like Random Forest and Logistic Regression instead of complex deep learning models. The system begins with data preprocessing to clean and prepare incoming data, after which the models identify unusual patterns that may indicate adversarial activity. By avoiding computationally heavy techniques, the approach remains efficient and practical. A user-friendly web interface built with Flask allows users to upload data and receive instant feedback. Overall, the project delivers an accessible, effective, and straightforward tool to enhance IoT security and defend against adversarial threats.