• AI-enabled Security Analytics: Adversarially Robust AI Agents, Cross-lingual Cybersecurity Analytics, Automatic Cyber Threat Identification
  • Statistical Machine Learning and AI: Adversarial Machine Learning, Transfer Learning and Domain Adaptation, Cross-lingual Knowledge Transfer, Reinforcement Learning, Deep Learning
  • Business Intelligence and Analytics: Social Media Analytics, Multilingual Product Review Analysis in E-commerce
  • Crime Data Mining: Online Predator Identification in Social Media, Supervised Methods for Categorizing Behavior of Offenders in Crime Incidents
  • Thesis Title: AI-enabled Cybersecurity with Transductive Learning, Transfer Learning, Adversarial Learning, and Reinforcement Learning Theory
  • Dissertation Summary: Cyber-attacks are a great societal concern. Many organizations rely on manual collection of cyber threat intelligence (CTI) to mitigate attacks. However, the fast-paced growth of data sources precludes obtaining actionable intelligence via manual approaches or ad-hoc software agents. AI-enabled cybersecurity is an emerging approach that draws upon statistical and machine learning theories to invent AI agents that address this issue. These agents can automatically gather CTI at a large scale and improve incident response. Although promising, these agents are vulnerable to adversarial attacks from AI-enabled adversaries. Given the crucial need for effective, robust cybersecurity AI agents, my dissertation presents five essays contributing to two major areas of AI-enabled cybersecurity: (1) AI-enabled cyber threat identification in international online hacker communities (three essays) and (2) Robustness of cybersecurity AI agents against adversarial attacks (two essays).