Intent-Aware Data Leak Mitigation for Contextual Data Sharing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current data leak prevention systems in financial institutions fail to understand the intent behind data transactions, leading to over-permissive or under-permissive data sharing, exposure of sensitive information, and manual, error-prone rule setups that are not scalable or adaptive.
Innovation Solution
An intent-aware data leak mitigation system using natural language processing (NLP) and semantic validation to understand the purpose of each data transaction in real-time, dynamically blocking or modifying data that does not align with intended business context, and maintaining an adaptive knowledge graph to refine data-sharing rules.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional data leak prevention systems are used, then data sharing can occur, but the systems cannot understand the intent behind data transactions leading to over-permissive or under-permissive data sharing
Solution Approach 1:
The patent introduces an intent detection intermediary layer between the data source and data destination. This intermediary uses natural language processing to analyze data requests, transaction contexts, and communication patterns to infer the intent behind data transactions. By inserting this mediator, the system gains intent understanding capability without fundamentally redesigning the entire data sharing infrastructure, thus improving measurement precision while managing device complexity.
Solution Approach 2:
The patent replaces traditional rule-based mechanical data leak prevention mechanisms with AI-driven intent detection systems. Instead of relying on static rules and patterns, the system uses machine learning models to dynamically understand and interpret the intent behind data transactions. This substitution enables more accurate intent detection while reducing the complexity of manually configuring and maintaining extensive rule sets.
2Adaptability or versatility
If manual rule setups are used for data leak prevention, then some level of protection is provided, but the rules are error-prone and not scalable
Solution Approach 1:
The patent implements self-service mechanisms where the intent detection system automatically learns from data transactions and continuously refines its understanding without manual intervention. The system autonomously analyzes transaction patterns, updates its models, and adapts to new types of data sharing scenarios. This self-service capability enables the system to maintain high adaptability while minimizing the need for manual rule configuration and updates, thereby increasing automation extent.
Solution Approach 2:
The patent incorporates feedback loops where the intent detection system continuously monitors data transactions, compares actual outcomes with expected results, and uses this feedback to refine its intent understanding. The system learns from both successful and unsuccessful data sharing attempts, adjusting its detection algorithms accordingly. This feedback mechanism enables the system to become increasingly adaptable over time while maintaining high levels of automation, as the system self-improves based on accumulated experience.
3Productivity
If data is shared without intent understanding, then data integration occurs, but sensitive information may be exposed
Solution Approach 1:
The patent applies preliminary action by performing intent detection and validation before data is actually shared. The system analyzes the requested data, determines the intent behind the request, validates whether the intent is legitimate, and only then permits data transmission. This pre-validation approach ensures that data security checks are completed in advance, allowing rapid data integration to proceed without compromising security, as the intent understanding framework is already in place to evaluate and approve or reject data sharing requests.
Data Source
AI summary
A data flow is intercepted, the intercepted data including a set of data fields carrying a corresponding set of content. An intent is identified behind a transaction including the intercepted data. A conformity indication is set relative to a first data field in the set of data fields to indicate nonconformity of a first content of the first data field, the setting including comparing the first content with expected data corresponding to a mapping of the intent in a knowledge graph. Based on the conformity indication, the first content is modified to form modified content. In the transaction in the data flow, the first content is replaced with the modified content, the replacing forming a modified transaction. The modified transaction is transmitted to a receiver data processing system over a data integration channel.