The fortnightly BRAG meeting will be held this Thursday, October 17th, at 1 pm via Zoom/GP-Y801. This week we will have presentations by Jamintha and Abdur.
Zoom Link: https://qut.zoom.us/j/89516271799?pwd=ajGvZnovJ6TPRW277NmywmbAqKARvu.1
Meeting ID: 895 1627 1799
Passcode: 723123
Jamintha’s Talk
Title: Hybrid Statistical Algorithmic Approach for Labour Market Delimitation
Abstract: In the intricate landscape of economic, demographic, and geographic factors, understanding the boundaries and characteristics of labour market areas (LMAs) has become a fundamental requirement for informed decision-making in policymaking and resource allocation. LMAs are functional regions representing areas where job seekers and employers interact to exchange labour services. Labor markets play a crucial role in determining employment patterns, regional development, and economic growth, making their accurate delineation essential for addressing regional disparities and guiding targeted interventions. Existing methods for labour market delimitation face significant challenges in accuracy and efficiency due to inherent limitations, including the failure to account for additional factors that help define labour markets, and the need for manual boundary adjustments. This research addresses these issues by developing an innovative, automated methodology that integrates multiple influential factors, offering a more precise and comprehensive analysis of labour markets and improving the overall accuracy of the delimitation process. The study begins with a comprehensive evaluation of the most contemporary and robust methodologies currently used to identify labour markets. These methodologies are then applied to simulated datasets, yielding distinct clusters that represent individual labour market areas. By critically examining the strengths and weaknesses of these methods, the study seeks to enhance the accuracy of labour market analysis and delimitation. The ultimate goal is to design a novel hybrid methodology that integrates statistical spatial modelling and simulated annealing, providing a more flexible and robust framework for labour market delimitation. This new approach will be better equipped to handle the complexities of modern labour markets and will be adaptable to evolving economic and demographic conditions, ensuring more effective and forward-looking policy decisions.
Abdur’s Talk
Title: Explainable fine-grained deep neural framework for multi-label software source code vulnerability discovery
Abstract: Software vulnerabilities are pieces of code that attackers can exploit, leading to various issues such as unauthorized access, privilege escalation, and remote code execution. The rise of vulnerable code, especially in open-source software, has resulted in financial losses, data breaches, and intellectual property infringements. The number of vulnerabilities has grown exponentially, with MITRE categorizing them under Common Weakness Enumerations (CWEs). Detecting vulnerabilities manually is resource-intensive, and with rapid software development, security is often overlooked. Automated detection methods, including pattern matching, machine learning, and deep learning, have been explored. Deep learning models, which use structural representations like abstract syntax tree, control/data flow and dependency graphs, etc., have shown potential for better generalization. However, existing methods are limited in identifying multiple co-occurring vulnerabilities, framing it as a multi-class problem instead of multi-label. Research on multi-label classification for software vulnerabilities is scarce, and a key challenge is the lack of suitable benchmarks. Additionally, there is a need for explainable AI models to improve interpretability and boost user confidence in vulnerability predictions. This research aims to develop a fine-grained explainable multi-label vulnerability discovery approach.