Botnet Attack Detection in IoT Using Machine Learning

Abstract

There are an increasing number of Internet of Things (IoT) devices connected to the network these days, and due to the advancement in technology, the security threads and cyberattacks, such as botnets, are emerging and evolving rapidly with high-risk attacks. These attacks disrupt IoT transition by disrupting networks and services for IoT devices. Many recent studies have proposed ML and DL techniques for detecting and classifying botnet attacks in the IoT environment. This study proposes machine learning methods for classifying binary classes. This purpose is served by using the publicly available dataset UNSW-NB15. This dataset resolved a class imbalance problem using the SMOTE-OverSampling technique. A complete machine learning pipeline was proposed, including exploratory data analysis, which provides detailed insights into the data, followed by preprocessing. During this process, the data passes through six fundamental steps. A decision tree, an XgBoost model, and a logistic regression model are proposed, trained, tested, and evaluated on the dataset. In addition to model accuracy, F1-score, recall, and precision are also considered. Based on all experiments, it is concluded that the decision tree outperformed with 94% test accuracy.

Enhancing the Security of Botnet Attacks Detection Using Parallel Gradient Descent Optimized Four Layered Network (PGDOFLN)

“…Decision Tree (DT) [20] struggles with biases in feature relevance in addition to vulnerable to overfitting, especially with complex datasets, and possibly lacking robustness in data with noise.…”

“…Although highly efficient, XgBoost [20] has difficulties such as high computational complexity and resource consumption, which could lead to longer training times, it is vulnerable to overfitting if hyperparameters aren't modified carefully. Despite its wide adoption, Logistic Regression (LR) [20] is frequently able to capture the complex patterns typical of botnet assaults because it presupposes linear decision bounds and is not able to describe complicated data relationships. The PGDOFLN has distinct advantages in terms of enhancing botnet detection and protection.…”

“…The effectiveness of the suggested approach and the existing approach was assessed with an "accuracy, precision, recall, and f1 score." The suggested method was compared with existing approaches, such as "Decision Tree (DT), Extreme Gradient Boosting (XgBoost), and logistic regression (LR)" [20].…”

Exploring a Hybrid Technique to Detect Network Intrusions in IoT Systems using PCC, ADASYN, and ALMA

“…In a method proposed by Alissa et al [2022], the decision tree achieved a test accuracy of 94% for binary classification. This method also achieved a precision, recall, and F1-score of 94%.…”

Artificial intelligence enabled Luong Attention and Hosmer Lemeshow Regression Window‐based attack detection in 6G

“…Gradient hybrid leader optimization (GHLBO) algorithm 23 and Hybrid Deep Learning method 24 were developed in Balasubramaniam et al 23 for discovering DDoS attacks. ML and DL techniques were investigated in Alissa et al 25 to determine and categorize botnet attacks. However, the large dataset was not focused.…”