Faculty of Computer Science and Information Technology, Universiti Malaya
Faculty of Computer Science and Information Technology, Universiti Malaya

AI and Cybersecurity Research

Home of the Bright, Land of the Brave

SIG Overview

The AI and Cybersecurity SIG stands at the forefront of the battle between artificial intelligence and modern digital threats, dedicated to securing the next generation of intelligent systems. Led by Dr. Hazim Hanif, this group bridges the gap between advanced machine learning and comprehensive cyber defense. Moving beyond traditional reactive measures, the group pioneers research into the dual nature of AI, by investigating both offensive AI tactics and intelligent threat detection across diverse environments. With deep expertise in adversarial machine learning and the security implications of Large Language Models (LLMs), the SIG focuses on developing resilient frameworks capable of anticipating, withstanding, and neutralizing sophisticated, AI-driven attacks across the digital landscape.

Key Research Areas

1
Large Language Models (LLMs) & Deep Representation Learning
Exploring the capabilities and limitations of transformer-based architectures for security applications, optimizing pre-training methodologies, and critically evaluating how data representations impact threat detection.
2
Adversarial Machine Learning & Model Robustness
Securing decentralized AI systems, particularly within federated learning and edge computing environments, by analyzing attack vectors and engineering robust defense mechanisms against adversarial manipulation.
3
Intelligent IoT Security & Network Traffic Analysis
Deploying AI-driven statistical analysis and transformer-based tokenization to accurately identify IoT devices and classify network traffic across diverse and emerging infrastructures, including 6G networks.
4
Automated Threat & Vulnerability Intelligence
Leveraging advanced machine learning, including hybrid Graph Neural Networks and deep learning ensembles to autonomously detect structural vulnerabilities and complex threats across various environments.
5
AI-Driven Digital Content Integrity
Utilizing supervised boosting models and deep learning frameworks for automated content moderation, detecting opinion spam, regional offensive text, and malicious profiles across digital platforms.

Research Projects

1
VulBERTa and the Efficacy of LLMs in Security
Pioneering simplified source code pre-training models (VulBERTa) and critically assessing the "Richer Representation Fallacy" to determine if complex LLM representations add value or noise to automated vulnerability detectors.
2
Deep Learning for Open-Source Intelligence and Sentiment Analytics
Utilizing stacked deep learning algorithms for large-scale data extraction, behavioral tracking, and sentiment analysis on social media platforms, demonstrating AI's capacity for complex digital intelligence gathering.
3
Comprehensive Taxonomies for AI-Assisted Vulnerability Detection
Systematically mapping the landscape of software vulnerabilities and evaluating the efficacy of various machine learning approaches to create foundational frameworks and benchmarks for future AI security tools.
4
Intelligent Moderation & Spam Detection Systems
Building tailored machine learning pipelines—ranging from supervised boosting for opinion spam detection to standardized Extra Tree models for regional offensive text classification (MOTEC), to preserve the integrity of digital communications.
5
Hybrid Graph Neural Networks for Code Security
Engineering hybrid GNN approaches to mathematically model complex execution paths, applied specifically to detecting sophisticated vulnerabilities in web-based applications and programming languages like PHP.