These are our current initiatives in AI, machine learning, and optimization.
Developing an efficient document classification framework, in collaboration with “Cyber Intelligence House (CIH)” company as the industry partner and supported by “DFCRC”, which is the Cooperative Research Centers program, an Australian Government initiative. We are trying to combine advanced language models with manually engineered structural features to detect and classify illicit online content across domains such as drug, weapon, and stolen credential marketplaces. The approach integrates NLP, ensemble learning, metaheuristic optimization, and robust machine learning models to address the challenges of heterogeneous, jargon-heavy, and obfuscated text from the deep/dark web, Telegram, Reddit, and Pastebin. Semi-supervised learning leverages unlabeled data for low-resource scenarios, while manual feature engineering captures structural layouts, embedded entities, and metadata to complement language model embeddings. Metaheuristics optimize feature selection and hyperparameters, resulting in a scalable, adaptable, and resilient solution for real-world cyber intelligence and online text classification.
Learn MoreOur previous research impact.
ARC DECRA Project "DE210101808" This project developed new tools and methods to improve how computers learn from data, especially when conditions change over time. It produced open-source software, including GenForge for evolving complex mathematical models, GNBG for generating customizable test problems, EDOLAB for teaching and testing algorithms in changing environments, and SPFP for splitting complex data into useful parts. These advances help researchers and engineers build smarter, more reliable systems for tasks like prediction and decision-making, while also supporting education and international collaboration in artificial intelligence.
Learn More