IB Multimedia Team

Investigating side-channel attacks on large language models. Supervised by Gururaj Saileshwar

Under the supervision of Professor Amir-massoud Farahmand and PhD candidate Tyler Kastner.

Investigated applications of distributional reinforcement learning to anomaly-based Intrusion Detection Systems (IDS).

• Developed a custom IDS environment using the NSL-KDD dataset

• Implemented Implicit Quantile Network (IQN), Quantile-Regression Deep Q-Network (QR-DQN), Categorical Q-Network (C-51) algorithms to train agents.

• Conducted literature reviews, analyzed results of experiments, drafted report and presented findings

Team: Threat Intelligence under Detection and Response

• Developed automated data pipelines and repositories for Indicator of Compromise (IOC) ingestion from vendors, APIs, emails, and form submissions

• Used regular expressions for IOC extraction and type-mapping as well as NLP models for text summarization.

• Automated scans of logs and internal data for the presence of found IOCs

• Designed and developed alerts to automatically cross-reference found IOCs for context and automatically generate a structured report of relevant information about threat actor and IOC.

• Designed and developed a dashboard of gathered insights to inform and visualize security and threat posture of Netflix.

Team: Clean Classification (machine learning focused cyber-security research team for Microsoft Defender)

• Developing datasets and training ML models to identify malicious software developers for Microsoft Defender and reducing false positives

• Data cleaning, normalization, feature engineering and anomaly detection in code signing certificates to classify malicious

developers

• Researching and developing a report on security vulnerabilities and indicators of compromised certificates