Skip to main content
Research Abstracts:
Smart Home attacks and detection
The Internet of Things (IoT) has become an integral part of our daily lives, contributing to the evolution of smart homes. However, the increasing prevalence of cyber-attacks targeting IoT devices has raised significant concerns among researchers and the general population, necessitating the development of robust security solutions. To enhance the detection of these attacks, it is crucial to establish effective modeling techniques to differentiate abnormal behavior from normal IoT device operations. This project addresses this challenge by developing two specific attack scenarios: man-in-the-middle and data injection attacks. By creating and analyzing these attack models, the student seeks to contribute to developing anomaly behavior analysis techniques that can enhance IoT systems' security. The insights gained from this project will aid in creating effective countermeasures to mitigate the risks associated with IoT-based attacks and safeguard the integrity and privacy of smart home environments.
Deep Learning for Network Intrusion Detection Systems
With the ubiquitous usage of computer networks and the plurality of applications running on them, cyber attackers attempt to exploit weak points of network architectures to steal, corrupt, or destroy valuable information. Network Intrusion Detection Systems (NIDSs) have become more significant in detecting novel attacks. Evaluating the efficiency of any NIDS requires a modern, comprehensive data set. The students will address these challenges using the UNSW-NB15 data set in this project. This data set includes nine categories of modern attack types and involves realistic activities of normal traffic captured with the change over time. In addition, it contains 49 features that comprise the flow based between hosts (i.e., client-to-server or server-to-client) and the packet header, which covers in-depth characteristics of the network traffic. The student will learn how to design the data science pipeline for machine learning-based NIDSs. The student will explore the procedures used for the data science pipeline, including the following steps: data collection, data preprocessing, data reduction, deep mode training and testing, and performance evaluation. Through the project, the student will develop the capability to analyze the network intrusion detection data set statistically and practically and use machine learning, particularly deep learning, to deal with NIDS tasks.
Intrusion Detection in Autonomous Vehicles
With the growth in autonomous vehicles, detecting and preventing attacks before malicious parties is imperative. This research focuses on detecting potential sensor attacks and anomalies in autonomous vehicles using an anomaly behavior analysis to identify sensor attacks. Real-world data was collected from a scalable autonomous vehicle testbed, including normal data representing regular vehicle behavior and attack data showcasing compromised sensor instances. This approach stands out from existing research by incorporating real-world attack data, offering greater realism and accuracy, and simplifying intrusion detection through a unified system capable of detecting a wide range of sensor attacks.
