LLM Log Data Masking for Sensitive Information Protection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional systems inadequately secure sensitive information in log files, exposing data such as SSNs, account numbers, and other personal information during log file generation and reporting.
Innovation Solution
A system utilizing a large language model (LLM) and generative artificial intelligence (GenAI) module to identify prone data, apply masking procedures, and secure sensitive information by replacing it with generalized messages or symbols, and transferring it to secured locations with permission-based access.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If conventional systems generate and report log files with sensitive information, then log files contain complete information for debugging and analysis, but sensitive information such as SSNs and account numbers is exposed
Solution Approach 1:
The patent extracts sensitive information from log files using a prone module that identifies and separates sensitive data elements. The GenAI module then applies masking procedures to remove or anonymize this extracted sensitive information, retaining only necessary non-sensitive details in the final log output. This resolves the contradiction by extracting harmful elements while preserving useful information.
Solution Approach 2:
The patent introduces a GenAI module as an intermediary between log generation and log storage/reporting. This intermediary applies intelligent masking procedures that selectively protect sensitive information while maintaining log utility. The DAO (Decentralized Autonomous Organization) acts as another intermediary layer to govern and approve masking policies, ensuring balanced protection without complete information loss.
2Reliability
If manual masking procedures are implemented to secure sensitive information, then data security is improved, but computing resources and time are consumed
Solution Approach 1:
The patent implements self-service masking where the GenAI module automatically identifies sensitive information patterns and applies appropriate masking procedures without manual intervention. The system trains on system-specific data to learn organization-specific sensitive information patterns, enabling autonomous operation that reduces both computing overhead and human resource requirements while maintaining high security reliability.
Solution Approach 2:
The patent performs preliminary training of the GenAI module using system-specific data, historical events, and existing permissions before deployment. This preliminary action prepares the system to automatically recognize and mask sensitive information efficiently during runtime, reducing the computational burden during actual log processing while ensuring accurate security protection.
3Reliability
If comprehensive masking procedures are applied to all data, then data security is maximized, but useful information for debugging and analysis is lost
Solution Approach 1:
The patent applies local quality masking where different masking strategies are applied to different portions of log data based on their sensitivity and utility. Critical sensitive information like SSNs and account numbers are fully masked or removed, while less sensitive contextual information is preserved. The GenAI module determines the appropriate masking level for each data element, ensuring security where needed and information retention where useful.
Solution Approach 2:
The patent implements partial masking actions rather than comprehensive masking of all data. The system identifies and masks only the specific sensitive information elements that require protection, leaving the rest of the log data intact and fully functional for debugging and analysis. This selective approach avoids excessive masking that would lose useful information while still providing adequate security protection.
Data Source
AI summary
Systems, computer program products, and methods are described herein for configuring data using advanced computational models for data analysis and automated processing. The present disclosure is configured to train a large language model (LLM), wherein training the LLM comprises using system-specific data comprising feed data, process run logs, historical events, code base, existing permissions, and data classification rules. The present disclosure is configured to determine prone data, wherein the prone data comprises a log file comprising sensitive information, and wherein the prone data is determined via a prone module. The present disclosure is configured to configure the prone data using a generative artificial intelligence (GenAI) module, wherein the GenAI module configures the prone data by masking the sensitive information using a masking procedure. The present disclosure is configured to determine the masking procedure via a decentralized autonomous organization (DAO).