Dilanka Kaushal Hewage

About

I received my MSc in Cybersecurity at the IU International University of Applied Sciences in Germany. Previously, I obtained my bachelor's degree from Singhania University. During my postgraduate studies, I was working on reasoning in LLMs for code generation and LLM-related cybersecurity; as a result, I developed a new approach for guiding the generation process. It consisted of a separate deep-think stage with self-reflection and a direct integration of thoughts (hidden states) into the LLM's main generation's hidden states using model surgery! This framework called LLM Reasoning Framework is applicable to various LLMs without finetuning. As an extension, It has a thought memory manager which controls the memories and thoughts of the LLM (not included in the evaluations) which allows the integration of approaches like Neural Turing Machines or DNC and LLMs.

Research Interests

My research focuses on the intersection of artificial intelligence and cybersecurity, particularly exploring how advanced reasoning capabilities in LLMs can be leveraged for security applications and threat detection.

A major challenge in current AI models is their struggle with generalization, consistent logical reasoning, and decision-making, especially in cybersecurity scenarios that require multi-step inference, vulnerability analysis, or handling previously unseen attack patterns. Current models (i.e., LLMs) can generate highly semantically coherent text, yet often fail in robust reasoning abilities necessary for tasks such as complex threat analysis, security planning, or adapting to novel cyber threats.

Researchers have explored many approaches such as prompt engineering methods (e.g., CoT and self-reflection), which can enhance reasoning abilities to an extent as a temporary and short-term solution; however, they are limited and are not architecturally native to the models. Therefore, my research focuses on developing differentiable approaches for implementing iterative reasoning and decision-making for LLM-powered security systems and autonomous threat detection agents. This involves developing new architectures that can maintain coherent thought processes across multiple inference steps in security-critical contexts.

Currently, I am exploring how advanced reasoning frameworks can improve LLM performance in code generation, vulnerability detection, and automated security analysis. My work involves integrating deep-think stages with self-reflection mechanisms and direct thought integration through model surgery techniques, enabling LLMs to reason more effectively about security problems without requiring extensive fine-tuning.

In a wider context, I am fascinated by the idea of developing intelligent security systems that combine specialized capabilities such as threat memory and rapid learning, goal-driven defense planning, anomaly detection, and error monitoring that operate collaboratively to make informed security decisions within complex environments. I am interested in transformers and attention mechanisms with applications in security contexts: adversarial robustness, multimodal threat detection, and self-supervised learning for zero-day vulnerability discovery.

News

• Completed MSc in Cybersecurity at IU International University of Applied Sciences, Germany - 2024
• Developed LLM Reasoning Framework for enhanced code generation and security analysis

Awards

• MSc in Cybersecurity, IU International University of Applied Sciences, Germany
• Bachelor's degree, Singhania University