Information Leakage Detection via Machine Learning Keyword Matching

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current methods for detecting information leakage in web applications are inefficient, as they fail to accurately identify sensitive information transmitted in plaintext due to variations in keyword matching, leading to increased labor and reduced efficiency for security testers.

Innovation Solution

A method using machine learning to match parameters in request messages against pre-identified keywords, determining a matching degree based on context, and recording information as leaked when the degree exceeds a threshold, thereby automatically identifying sensitive information more efficiently.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional keyword matching method is used to detect sensitive information, then the detection process is simple, but the detection accuracy is low and many sensitive information types are missed

Engineering Contradiction:
Improvedetection accuracyVSAvoiddetection method complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent changes the detection parameters from simple exact keyword matching to multi-dimensional parameter matching including fuzzy matching, case-insensitive matching, and context-based matching. This allows the system to detect variations of sensitive information keywords while maintaining a manageable detection framework.

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The patent segments the detection process into multiple stages: keyword extraction from training data, parameter matching against sensitive information patterns, context analysis, and final detection judgment. This segmentation improves accuracy while keeping each stage relatively simple.

Inventive Principle:
Principle #1Segmentation

2Productivity

If traditional keyword matching method is used, then the detection method is easy to implement, but the productivity of security testing is reduced due to manual analysis requirements

Engineering Contradiction:
Improvesecurity testing efficiencyVSAvoiddetection method ease of use
Core Design Contradiction:
ProductivityVSEase of operation

Solution Approach 1:

The patent implements self-service through automated keyword extraction from training data and automatic parameter matching against detection patterns. The system autonomously performs detection without requiring manual security analyst intervention for each check, thereby improving productivity while maintaining ease of operation through centralized configuration.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The patent performs preliminary actions by pre-training the detection model with labeled data to extract sensitive information keywords and establish detection patterns before actual security testing. This preliminary preparation enables automated, high-productivity detection during the actual testing phase.

Inventive Principle:
Principle #10Preliminary action

3Reliability

If simple keyword matching is used, then the detection system is fast and efficient, but the reliability of detection results is low due to false negatives

Engineering Contradiction:
Improvedetection reliabilityVSAvoiddetection time
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The patent applies partial matching rather than requiring exact matches, using fuzzy matching and pattern recognition to identify sensitive information even when keywords appear in varied forms. This increases reliability by reducing false negatives while maintaining reasonable detection speed through efficient pattern matching algorithms.

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentEP3447669B1Information leakage detection method and device, server, and computer-readable storage medium
Publication Date: 2020.11.11 PING AN TECH (SHENZHEN) CO LTD
  • EP3447669B1 patent drawingFigure 1~2
  • EP3447669B1 patent drawingFigure 3
  • EP3447669B1 patent drawingFigure 4~5

AI summary

Disclosed is a method for detecting information leakage that includes: receiving a request message sent from a channel party; matching a parameter included in the request message against a keyword obtained in advance through machine learning and obtaining a matching degree between the parameter and the keyword based on the matching result; and determining information of the request message is leaked and recording related information of the request message when the matching degree is higher than a preset threshold. Also disclosed are a device for detecting information leakage, a server, and a computer-readable storage medium. Based on machine learning, more keywords can be automatically identified, allowing the sensitive information to be identified in very efficient manner, leading to reduced efforts required of security testers, improved efficiency of detecting information leakage and reduced risk of information leakage.