A Lightweight Edge-Computing Framework for IoT Attack Detection Under Resource Constraints

This preliminary research investigates machine learning approaches for detecting IoT network attacks under resource constraints, focusing on distinguishing benign traffic from UDP flood attack traffic. The study evaluates Decision Tree, Logistic Regression, and Random Forest classifiers using network traffic data. It also examines the effect of removing the Protocol Type feature and investigates duplicate records and data leakage risks. The results show strong classification performance on the selected dataset. However, the findings should be interpreted cautiously because the evaluation uses traffic from only two source files and does not establish generalization to independent network captures, other attack types, or real edge devices. This work is a preliminary research study and has not been peer reviewed.