Document-to-Template Matching for Data Leak Prevention
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current DLP methods face challenges in efficiently classifying and identifying outbound documents, including the need for storing large amounts of data, potential exposure of restricted information, and security vulnerabilities due to incomplete fingerprinting mechanisms.
Innovation Solution
The method involves tokenizing documents into serialized lines, calculating hash values, and matching them against a centralized template set to determine similarity, allowing for real-time classification and enforcement of security policies without storing entire documents, using a hash map to efficiently compare documents to multiple templates.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If brute-force fingerprinting of the whole corporate data is used to classify each document, then document classification accuracy is improved, but storage requirements and processing time increase significantly
Solution Approach 1:
The patent segments the document classification problem by dividing documents into structured sections (header, body, footer) and comparing only corresponding sections against template documents. This selective section-by-section comparison approach maintains classification accuracy while dramatically reducing the amount of data that needs to be stored and processed, as opposed to fingerprinting the entire document corpus.
Solution Approach 2:
The patent extracts and compares only the essential structural components of documents (header, body, footer sections) rather than storing or processing the complete document content. This extraction of key structural elements enables accurate template matching while minimizing storage requirements and processing overhead.
2Measurement precision
If brute-force fingerprinting of the whole corporate data is used to classify each document, then document classification accuracy is improved, but processing time increases significantly
Solution Approach 1:
The patent segments the document comparison process into discrete structural sections (header, body, footer) that can be independently compared against corresponding template sections. This segmentation enables parallel processing of different document sections and eliminates the need to process entire documents sequentially, significantly reducing processing time while maintaining classification accuracy.
Solution Approach 2:
The patent performs preliminary actions by pre-defining template documents with their structural sections (header, body, footer) before actual document classification occurs. This pre-preparation of templates with expected structural elements allows for rapid comparison against new documents without requiring comprehensive fingerprinting of the entire corporate document database during processing.
3Adaptability or versatility
If centralized database storage of corporate data is implemented, then document classification capability is improved, but security risks increase due to mass exposure of restricted data
Solution Approach 1:
The patent extracts only the essential structural template information (header, body, footer patterns) from corporate documents and stores these templates centrally, rather than storing complete document content. This extraction approach enables document classification capability while minimizing security risks, as the centralized database contains only structural patterns rather than sensitive document data.
Solution Approach 2:
The patent creates simplified copies of document structures in the form of templates that capture the essential classification characteristics without containing the actual sensitive content. These template copies enable classification functionality while the original restricted data remains secure and is not mass-stored in the centralized database.
4Reliability
If fingerprinting mechanism is restricted for security reasons, then security vulnerabilities are reduced, but document classification accuracy deteriorates
Solution Approach 1:
The patent segments the document into structural sections (header, body, footer) and applies fingerprinting only to these essential structural elements rather than the entire document. This selective fingerprinting approach maintains document classification accuracy by capturing the characteristic structure of different document types while restricting access to and processing of sensitive content, thereby maintaining both security reliability and classification precision.
Data Source
AI summary
The present invention discloses methods for document-to-template matching for data-leak prevention (DLP), the methods including the steps of: providing a document as a stream of characters; splitting the stream into a plurality of serialized data lines; calculating a hash value for each serialized data line; checking for each hash value in a hash map of a template set; determining a similarity match to a particular template based on a predefined threshold of template hash values, of the template set, being found in the stream; and based on the similarity match, executing a DLP security policy for the document. Preferably, the template set is extracted from documents manually prepared by a security administrator. Preferably, each template in the template set is deduced automatically from a plurality of documents.