Description:
UTHealth creators have developed AI SpamShield, a standalone AI spam-purging tool that could easily be integrated into any IMAP-based email system. AI SpamShield’s capabilities include skilled distinction between legitimate and spam messages, complemented by the abiliity to continuously learn from user feedback that further enhances accuracy for individual users over time. Additional features such as automatic folder allocation can boost business productivity and efficiency.
Background
Traditional spam filters for email messages perform marginally because they rely on predefined rules and static blacklists, thus lacking the ability to automatically adopt to evolving spam tactics.
Significance and Impact
UTHealth researchers developed AI SpamShield which is a standalone, IMAP-compatible, AI-driven spam-purging system. AI SpamShield classifies email messages as spam or non-spam based on AI-developed guidance, supplemented by user tagging. Accuracy of the system for individual users thus improves over time based on their feedback, creating a context-aware personalized assistant to detect spams. The system can also prioritize message and allocate them to user-specified folders. AI SpamShield’s unique capabilities can therefore improve personal and business efficiency.
Benefits/Technology Advantages
• Standalone software that is compatible with any email system that supports IMAP.
• Enhanced adaptivity to diverse spam strategies: Using machine learning to dynamically learn from user interactions.
• Continual, adversarial learning: Nightly self-training detects shifting spam tactics, improving accuracy silently.
• Context-rich triage: Outputs real-time risk scores (based on similarity) and reasoning (based on agent analysis) via API, beyond simple spam/not-spam.
• Zero-friction deployment: IMAP daemon installs in < 15 min, is language-agnostic, and requires no MX changes or lock-in.
• Improved business efficiency with email organization and prioritization.
Intellectual Property Status
Under development and available for licensing.
About the Inventors
Daniel Sessler, M.D.
Professor of Anesthesiology, Critical Care, and Pain Medicine and VP, Clinical and Outcomes Research at UTHealth Houston
Xiaoqian Jiang, Ph.D.
Associate VP for Medical AI Chair, Department of Health Data Science and AI at UTHealth Houston