Sparse 0 Mod P Fingerprinting for Leakage Prevention
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current information leakage prevention systems face inefficiencies in document fingerprinting, particularly with classical 0 mod P and Winnowing techniques, which result in high data storage needs and lack of guaranteed match features for long content fractions.
Innovation Solution
The Sparse 0 Mod P technique reduces fingerprint data size while maintaining a guaranteed minimum match size by selecting only one 0 mod P hash value per window and using Bloom filters for efficient matching, significantly reducing data storage and processing requirements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If classical 0 mod P technique is used for document fingerprinting, then match accuracy is improved, but data storage requirements increase
Solution Approach 1:
The patent extracts only the essential fingerprint information needed for matching by using Bloom filters, which store compact bit representations of document fingerprints. This extraction principle reduces the fingerprint data size from storing complete hash values to storing only critical presence/absence information in a condensed filter structure.
Solution Approach 2:
Instead of storing complete fingerprint data and searching through it, the patent inverts the approach by using Bloom filters to store presence information and check for matches efficiently. The inversion transforms the problem from storing and searching large fingerprint datasets to using compact filter structures that enable rapid match verification with minimal storage.
2Productivity
If Winnowing technique is used for document fingerprinting, then processing speed is improved, but match reliability for long content fractions deteriorates
Solution Approach 1:
The patent changes the parameters of the fingerprinting system by using Bloom filters with configurable hash functions and bit array sizes. This allows optimization of both processing speed and match reliability by adjusting filter parameters such as the number of hash functions and filter capacity, enabling the system to maintain high processing speeds while guaranteeing matches for long content fractions.
3Reliability
If more fingerprints are selected per document, then match accuracy is improved, but data storage and processing requirements increase
Solution Approach 1:
The patent extracts only the essential presence/absence information of fingerprints using Bloom filters, discarding redundant detailed fingerprint data. This extraction maintains match accuracy by preserving the critical information needed for matching while eliminating unnecessary data that would increase processing complexity.
Solution Approach 2:
The Bloom filter structure acts as a disposable, lightweight data structure that can be quickly created and queried. Instead of maintaining complex, long-lived fingerprint databases requiring extensive processing, the system uses these simple, easily manageable filter structures that provide sufficient match accuracy with minimal processing overhead.
Data Source
AI summary
Processes for fingerprinting a document and for preventing information leakage at a deployment point are disclosed. For fingerprinting a document, a sequence of hash values for a document is generated, a portion of said hash values to be selected as fingerprints for the document. A current window is positioned over a portion of the sequence of hash values. The hash values are examined starting from one end of the current window, and a first-encountered hash value that is 0 modulo P is selected to be a fingerprint for the current window. For information leakage prevention at a deployment point, a rolling hash calculation is performed on a target document, and a determination is made if a hash value is 0 modulo P. A first filter is applied if the hash value is 0 modulo P, and a second filter is otherwise applied. Other embodiments, aspects and features are also disclosed.