Intelligent Data Security Engine for Internal Breach Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing computer systems are limited in their ability to control what can be accessed by particular groups of users, leading to internal data breaches, which can cause accidental or unlawful destruction, loss, or unauthorized disclosure of sensitive information.
Innovation Solution
A system utilizing intelligent data information security engine with machine learning models, including self-organizing maps, LSTM networks, and random forest algorithms, to analyze user activity data and query intent, determining data breach fields and ranking values to prevent unauthorized access.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional access control systems are used to manage user data access, then system simplicity is maintained, but the ability to detect and prevent internal data breaches is insufficient
Solution Approach 1:
The system performs preliminary analysis of user activity data before data access requests are executed. Machine learning models are trained in advance on historical user behavior patterns to establish baseline expectations. When a data access request occurs, the pre-trained models can immediately evaluate whether the request aligns with normal user behavior, enabling proactive breach detection rather than reactive response.
Solution Approach 2:
The patent introduces an intermediary security system that sits between users and the data they wish to access. This intermediary layer includes multiple machine learning models (SOM for clustering, LSTM for sequence prediction, random forest for classification) that analyze user behavior patterns and mediate access decisions. This intermediary acts as a intelligent gatekeeper that enhances security without requiring fundamental changes to the underlying data storage or access infrastructure.
2Measurement precision
If machine learning models are deployed to analyze user behavior patterns, then data breach detection accuracy is improved, but computational processing time increases
Solution Approach 1:
The patent segments the machine learning analysis into distinct modular components: SOM models handle clustering of user behavior patterns, LSTM models handle sequential behavior analysis, and random forest models handle classification of access requests. Each model processes specific aspects of user behavior independently, allowing parallel execution and reducing overall processing time while maintaining comprehensive analysis accuracy.
Solution Approach 2:
The system implements partial analysis by focusing machine learning models only on specific high-risk data access patterns rather than analyzing all user activities equally. The SOM models identify clusters of concerning behaviors, and subsequent LSTM and random forest analyses are applied selectively to requests within those clusters. This partial action approach reduces computational overhead while maintaining high detection accuracy for critical breach scenarios.
Data Source
AI summary
A system for receiving, from a plurality of systems, data breach criteria and user activity data associated with a user. The system uses a first machine learning model, the data breach criteria, and the user activity data to determine a plurality of data breach fields and information values for the plurality of data breach fields. The system receives a query from the user to request access to a dataset associated with the plurality of systems. The system uses a second machine learning model to determine data sensitive fields in the query. The system uses a third machine learning model to determine ranking values for the data sensitive fields and a data information loss value. In response to determining the data information loss value is smaller than a predetermined threshold, the system validates the query from the user to allow the user to access the dataset.