International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 12 Issue: 02 | Feb 2025
p-ISSN: 2395-0072
www.irjet.net
Empowering IOT Cyber Network Attacks using Machine Learning Mrs. J Preethi Lakshmi1, Dr. T Aravind2, Dr S. Praveenkumar3, Mr. T.K Srinivasan4 1M.E Applied Electronics, Department of ECE, Saveetha Engineering College, Thandalam, Chennai,
Tamil Nadu, India.
2 Professor, Department of ECE, Saveetha Engineering College, Thandalam, Chennai, Tamil Nadu, India. 3 Professor, Department of ECE, Saveetha Engineering College, Thandalam, Chennai, Tamil Nadu, India. 4Professor, Department of ECE, Saveetha Engineering College, Thandalam, Chennai, Tamil Nadu, India. ---------------------------------------------------------------------***--------------------------------------------------------------------Abstract—Advanced techniques for detection and mitigation of cyber-attacks on IoT systems are in high demand due to their
increasing prevalence. This work deals with the challenge by using machine learning algorithms, XGBoost and LightGBM, for classification of network traffic patterns based on the NSL-KDD dataset. The main problem here is the identification of different cyber-attacks, such as DoS, DDoS, and probing attacks, in real-time within IoT networks. The solution proposed incorporates XGBoost and LightGBM, which can be done to enhance classification accuracy to the maximum level and improve the scalability of the system while ensuring malware is effectively detected. The timeliness of response by the system is ensured by classifying network traffic as benign or malicious. The advantages of this approach include high detection accuracy, scalability, and enhanced security for the IoT infrastructures while ensuring robust protection against many cyber-attacks.
Keywords: IoT Security, Network Traffic Analysis, Cyber Threat Detection, Real-Time Monitoring, Malicious Activity, Scalable Protection, Attack Classification.
I.
INTRODUCTION
The Internet of Things has revolutionized the way we interact with technology, connecting a wide variety of devices and enabling smarter environments. From smart homes to industrial automation, IoT systems have become part of daily life. However, the growth of IoT networks has also brought along significant security challenges. The nature of these systems depends on continuous connectivity; thus, they are prone to different forms of cyber-attacks. The sophistication of cyber threats targeting IoT devices and networks [1] has emphasized the requirement for advanced security mechanisms that can detect and mitigate the risks in real-time. A successful cyber-attack on an IoT system may have consequences ranging from a mere data breach to disruption of services or even hijacking of connected devices. Thus, ensuring the security and integrity of IoT systems is one of the critical factors in determining their continued success and adoption. The forms in which cyber-attacks targeting IoT networks come include Denial of Service (DoS), Distributed Denial of Service (DDoS), and probing attacks. DoS and DDoS attacks aim at overwhelming IoT devices or networks by flooding them with excessive traffic, rendering them unusable. Probing attacks involve scanning and probing network devices for vulnerabilities that can be exploited by attackers. Identifying these malicious [2] activities in real-time is important to minimize their impact and ensure the smooth operation of IoT systems. Traditional security mechanisms, such as firewalls and intrusion detection systems (IDS), are usually not able to deal with the dynamic nature and diversity of IoT traffic. Due to the variety of devices, network protocols, and communication patterns in IoT systems, developing a one-size-fits-all security solution is challenging. Hence, there is an urgent need for adaptive and intelligent security techniques that can detect emerging threats and respond effectively to them. Machine learning has been one of the promising solutions for these limitations of traditional security systems. Machine learning algorithms can analyze large volumes of network traffic data, learning to distinguish between benign and malicious activity through the identification of patterns and anomalies. Among these machine learning techniques, ensemble methods such as XGBoost and LightGBM have been widely used for classification tasks because they are able to handle complex and high-dimensional data. These algorithms can be quite appropriate for cybersecurity applications, as they can identify subtle patterns in network traffic that may indicate cyber-attacks. XGBoost is highly performant and scalable, while LightGBM is efficient in handling [3] large datasets; thus, both are good choices for dealing with the challenge of IoT security. This work presents a method for improving the detection and mitigation of cyber-attacks in IoT systems by combining XGBoost and LightGBM algorithms for network traffic classification. The proposed methodology is trained and validated on the well-known NSL-KDD dataset, widely used to train and evaluate network intrusion detection systems. The dataset contains labeled network traffic data, including normal and attack traffic, which can be used to teach models to differentiate between benign and malicious activities. By training these algorithms with this dataset, the system will learn how to classify traffic into different categories, such as DoS,
© 2025, IRJET
|
Impact Factor value: 8.315
|
ISO 9001:2008 Certified Journal
|
Page 543