Content Sensitivity Classification via Partial Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current technologies lack an efficient method to monitor and prevent the misuse or theft of sensitive information, such as financial data and intellectual property, within corporate networks, as existing access control systems fail to track and control the transmission of sensitive content effectively, leading to potential legal and financial repercussions.
Innovation Solution
A content monitoring system, comprising a content appliance that uses linguistic analysis and key repositories to classify and track sensitive information, allowing for real-time detection and prevention of unauthorized transmission, even of partial document excerpts, without altering existing business processes or infrastructure.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If keyword or key-phrase matching is used for content monitoring, then sensitive information can be identified, but false positives occur and overly restrictive blocking results
Solution Approach 1:
The patent segments content monitoring into multiple levels: document-level classification using access control lists, section-level sensitivity analysis, and passage-level keyword matching. This hierarchical segmentation allows the system to apply appropriate monitoring intensity to different content granularities, reducing false positives while maintaining detection accuracy for actual sensitive information.
Solution Approach 2:
The system applies different monitoring strategies to different portions of content based on their sensitivity characteristics. Access control lists provide coarse-grained classification for entire documents, while keyword matching provides fine-grained detection for specific sensitive passages. This local quality differentiation enables precise control without blanket restrictions.
2Ease of operation
If access control lists are used to control document access, then access permissions can be managed, but transmission control after access is granted cannot be achieved
Solution Approach 1:
The system implements feedback by continuously monitoring outgoing transmissions and comparing them against classified sensitive information. When matches are detected, the system provides feedback control by blocking or alerting on attempted transmissions. This closed-loop feedback mechanism bridges the gap between access control and transmission control, preventing information loss despite granted access permissions.
3Loss of time
If tracking information movement by filename or URL is used, then transmission paths can be monitored, but content-based protection cannot be achieved
Solution Approach 1:
The patent merges metadata-based tracking (filenames, URLs, access control lists) with content-based analysis (keyword matching, section classification). This combination allows the system to use efficient metadata for initial filtering and routing, then apply precise content analysis only where needed, achieving both speed and accuracy in sensitive information protection.
4Reliability
If blanket blocking of sensitive material transmission is implemented, then all sensitive information can be protected, but legitimate business communications are overly restricted
Solution Approach 1:
The system dynamically adjusts transmission control based on real-time analysis of content sensitivity and recipient context. Rather than static blanket blocking, the system evaluates each transmission attempt individually, allowing legitimate business communications to proceed while blocking only those containing actual sensitive information. This dynamic approach maintains both protection reliability and communication productivity.
Data Source
AI summary
Monitored content is analyzed to determine full and partial matches to previously classified content. Monitored content matching previously classified public content is classified as public, even if the monitored content is also found to match previously classified private content. In other words, public classification “overrides” potentially private classification. Monitored content matching only previously classified private content is classified as private. All remaining otherwise unclassified monitored content is classified as unknown. Monitored content is analyzed with respect to a session. If any content in a session is private, then the session is classified as private. If all content in a session is public, then the session is classified as public. Otherwise, the session is classified as unknown. In a related aspect, a set of policies are searched for a first match in part according to the classification, and a designated action taken if the first match is found.