KGC 2024: Dowsing for Data: A Unified Knowledge Model for Data Loss Prevention
Please enjoy this talk “Dowsing for Data: A Unified Knowledge Model for Data Loss Prevention,” from the Knowledge Graph Conference 2024 given by Michael Miller and Zach Tretter from Northrop Grumman.
Talk description:
A scalable knowledge protection strategy is critical to an organization’s security and success. Companies safeguard sensitive metrics, intellectual property, and information regulated by governments.
Typically, protection is implemented at two levels: High-level guidance about restricted topics is provided by organization leadership, customers, or regulatory bodies. Then, policy enforcement occurs at the token level. An author or algorithm must determine whether a combination of words or paragraphs is in violation of a high-level policy.
The gap between policy and enforcement is context-sensitive and often requires expert understanding. Even with manual review and automated methods, errors are frequent and have the potential to be catastrophic.
We propose Dowser, a holistic approach to Data Loss Prevention (DLP). Dowser is an interactive knowledge graph that captures topic-level guidance and token-level enforcement in the same database. This approach has several benefits: an explainable, maintainable model of high-level guidance, and a robust framework for policy enforcement.
Dowser can be implemented across security environments, and it can interface with Large Language Models (LLMs) for state-of-the-art automation. Ultimately, Dowser enhances knowledge security by making DLP more transparent for implementers, reviewers, and regulators.
Technical topics covered:
- Knowledge Management
- Data Modeling
- Knowledge Representation
- Natural Language Processing (NLP)
- Knowledge Graph Reasoning
- Knowledge Graph Applications
Do you enjoy talks like this one? Subscribe to the KGC newsletter and stay up to date with everything happening in the world of knowledge graphs, AI, and semantic tech.
***
