Machine Learning Sensitive Data Detection System
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing sensitive information detection software is ineffective in identifying and preventing sensitive information from being stored due to reliance on regex patterns, leading to false positives and ease of circumvention by altering data formats, resulting in organizations inadvertently storing excessive sensitive information.
Innovation Solution
A method using a natural language understanding model and machine learning to detect sensitive information entered into a form on a user interface, generating comprehension data and determining a risk score, with alerts and actions taken based on the score to manage and store sensitive information effectively.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If regex patterns are used for sensitive information detection, then the detection process is simple and fast, but the accuracy is low and false positives are frequent
Solution Approach 1:
The patent replaces the mechanical regex pattern matching system with a machine learning-based detection system. The machine learning model processes input data through neural networks to identify sensitive information, substituting the rigid mechanical regex approach with an adaptive intelligent system that achieves both speed and accuracy.
Solution Approach 2:
The patent changes the detection parameters from fixed regex patterns to dynamic machine learning models that can adapt to different data formats. The model learns from training data and adjusts its detection criteria, transforming the static parameter-based detection into a dynamic, learning-based approach that improves accuracy while maintaining efficiency.
2Ease of manufacture
If regex patterns are used for sensitive information detection, then the implementation is straightforward, but the system is easily circumvented by altering data formats
Solution Approach 1:
The patent introduces dynamics into the detection system by using machine learning models that can adapt to different data formats and expressions. Unlike static regex patterns, the machine learning model continuously learns from data and can recognize sensitive information regardless of how it is formatted, making the system dynamic and resilient to circumvention attempts.
Solution Approach 2:
The patent implements feedback mechanisms where the system learns from detected patterns and adjusts its detection criteria over time. The machine learning model uses feedback from training data and can be fine-tuned based on new information, creating a self-improving system that becomes increasingly robust against circumvention strategies.
3Device complexity
If existing detection software is used, then the initial setup is simple, but the false positives cause users to ignore alerts
Solution Approach 1:
The patent replaces the simple but inaccurate regex-based system with a machine learning system that provides more reliable alerts. By substituting the mechanical pattern matching with intelligent models, the system achieves higher alert accuracy while maintaining reasonable complexity through automated learning processes.
Data Source
AI summary
The disclosed technology includes systems and methods of identifying and managing sensitive information. The method can include detecting data entered into a form on a user interface, and generating, by a natural language understanding model, comprehension data based on the data entered into the form. The method can include determining, by a machine learning model, a risk score of the data and determining whether the risk score is greater than or equal to a threshold risk score. In response to determining that the risk score is greater than or equal to the threshold risk score, the method can include outputting a message to the user interface to alert a user that the data likely comprises sensitive information. The method can include receiving an input from the user interface indicative of whether the data comprises sensitive information and storing the data, the risk score, and the input in a database.