Privacy Document Analysis
In an era where data privacy is at the forefront of public and regulatory concerns, privacy policies serve as a key mechanism for organizations to communicate their data handling practices. However, these documents are often long, complex, and difficult for users, regulators, and even legal professionals to interpret. At the same time, advancements in AI and NLP have presented the possibility of automatically extracting meaningful and useful information from complex documents, and ease the difficulty of comprehension. We are exploring multiple NLP techniques to extract granular data from privacy policy documents, with the objective of reorganizing the data into more presentable formats such as summary documents, question answering, and visual notices, in an automated manner. The extracted data can also serve novel analytics on privacy policy documents.

LLMs and Privacy Comprehension
Large Language Models (LLMs) have shown remarkable capabilities in understanding and generating human language. However, when it comes to comprehending privacy documents, LLMs may encounter interpretation gaps due to the complex and nuanced nature of legal language. Our research focuses on identifying these gaps, characterizing them based on their source and impact, and measuring their prevalence across different types of privacy documents. By understanding these limitations, we aim to develop methods to improve LLM comprehension of privacy documents and ensure that individuals' privacy rights are adequately protected in the age of AI.

Situational Awareness in the Power Grid
The power system is one of the most critical infrastructures in modern society. It has been a target for cyber and physical attacks in the past two decades. With different attacks that have occurred and have been recognized, anomaly or intrusion detection systems are now in demand in the power domain. We are exploring research questions on how we can work towards a balanced tolerance of a neural network model that is able to achieve high performance metrics while leaving little to no avenues for an adversary to perform a successful attack. While making changes and improvements to the detection model, we are also addressing the challenges of applying such a real-time model into the real world and at the same time assess a model's operability with future technologies that involve edge computing.

Real-Time Log Analysis for PII
The increased amount of web applications and internet software solutions utilizing cloud frameworks has contributed to large data sets of system and developer log messages being generated constantly. These messages may contain sensitive data, creating an additional security risk for the systems, and contributing to the need for analysis of such large volumes of data in real time. Large commercial data monitoring systems can solve these analysis requirements, but they can be costly. We are exploring open source solutions, and enterprise scale deployment options, to analyze such log data, which ingests it, processes it and visualizes sensitive data found within in real time.