She is currently a PhD candidate in the collaborative program between BITS Pilani and RMIT University. Her research focuses on enhancing the robustness of foundation models through adversarial deep learning. Through her research, she will be exploring innovative techniques to identify and mitigate various adversarial scenarios, particularly in applications spanning financial markets, business scenarios, social media to name a few. Through her work, she aims to contribute to delivering impactful insights that addresses real-world challenges and pave the way for more resilient technological advancements across various sectors.

PhD Research Project

Adversarial Deep Learning Algorithms Applied to Outlier Detection in Dynamic Networks

This research study proposes to develop novel adversarial deep learning methods for unsupervised scenarios like outlier detection in dynamic networks with labeled data. It will study the coevolution of homogeneous and heterogeneous dynamic networks to identify subgraph anomalies in weakly-supervised, sparse, linked and streaming datasets found in intelligent systems. With the help of loss functions, the anomalies will be scored, and continual lifelong learning approaches will be proposed that incorporate class and cost distribution information for feature representations. The project will leverage explainable AI to interpret deep learning based outlier detection, address adversarial machine learning challenges, and develop cost-sensitive classification algorithms for unequal and unknown misclassification costs. The research aims to advance the state-of-the-art in outlier detection for dynamic networks by combining deep generative models, adversarial training, and explainable AI on challenging real-world datasets.

Supervision

• Dr. Yee Ling Boo
• Dr. Aneesh Chivukula
• Dr. Araz Nasirian
• Dr. Yiliao Song
• Dr. Apurba Das