What type of study is this?

September 10, 2025Open Access

Enhancing IoT Security: A Comparative Analysis of Machine Learning and Deep Learning Techniques for Botnet Detection

Key Points

Data augmentation significantly enhances model performance in detecting IoT threats, improving precision and recall.
Both decision tree and k-nearest neighbors achieved top metrics with an accuracy of 0.98 in detecting Mirai botnet attacks.
The study utilized the cic iot dataset 2023 to evaluate various ML and DL models for IoT security applications.
Findings indicate that balanced datasets are crucial, as SGD's performance suffered due to class imbalance.

Abstract

The Internet of Things (IoT) has revolutionized technological interactions but still faces significant security challenges from threats such as botnets. Therefore, effective detection methods are crucial. This study evaluates several Machine Learning (ML) and Deep Learning (DL) models for detecting IoT cyber threats, focusing on Mirai botnet attacks and ARP spoofing on the CIC IoT Dataset 2023. ML models, namely Stochastic Gradient Descent (SGD), Support Vector Machine (SVM), Decision Tree (DT), Logistic Regression (LR), and K-Nearest Neighbors (KNN), and DL techniques, namely Feedforward Neural Network (FNN) and Convolutional Neural Network (CNN), were evaluated. The results show that data augmentation (oversampling) significantly increased performance across all models. DT and KNN achieved the highest metrics (precision, recall, F1-score, and accuracy of 0.98), demonstrating superior classification capabilities. DL models had similar results, with CNN improving from 0.96 to 0.98 after oversampling, showing its adaptability to enhanced data diversity. Conversely, SGD demonstrated high sensitivity to class imbalance, emphasizing the need for balanced datasets in IoT security applications.

Read Full Paperexternally

KI fragen

Bookmark

View Full Paper