Synthetic Test Data Generation for Privacy-Safe Application Testing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing software testing methods face challenges in accurately assessing application performance without exposing protected or private data, leading to potential data leaks and non-representative test results due to the exclusion of such data.
Innovation Solution
A system utilizing machine learning models to generate artificial test data that shares characteristics with original protected data, allowing for secure and representative testing without exposing the original data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If private data is excluded from test datasets, then data security and privacy protection are improved, but test representativeness and accuracy deteriorate
Solution Approach 1:
The patent generates synthetic test data that copies the statistical characteristics, distributions, and relationships of private data without containing actual private information. Machine learning models learn patterns from private data and generate synthetic copies that preserve test representativeness while eliminating privacy risks.
Solution Approach 2:
The patent transforms private data by changing its parameters and characteristics through synthetic data generation. The synthetic data maintains essential statistical properties and relationships while altering specific values and identities, thereby preserving test accuracy without exposing private information.
2Measurement precision
If private data is used in test datasets, then test accuracy and representativeness are improved, but data leak risks and privacy violations increase
Solution Approach 1:
The system creates synthetic copies of private data that capture essential patterns and relationships needed for accurate testing. These copies enable comprehensive test coverage while eliminating the risk of actual data leaks, as the synthetic data cannot be traced back to specific private records.
Solution Approach 2:
The patent introduces synthetic data as an intermediary between private data and testing processes. This intermediary preserves the beneficial statistical properties for test accuracy while blocking the harmful data leak pathway, allowing testing to proceed without direct access to sensitive information.
3Reliability
If synthetic data is generated to replace private data, then data privacy is protected, but data processing complexity increases
Solution Approach 1:
The patent performs synthetic data generation in advance, creating ready-to-use test datasets before actual testing begins. This preliminary action eliminates the need for complex real-time data processing during testing, as the synthetic data is pre-generated with all necessary characteristics embedded.
Solution Approach 2:
The system replaces complex manual data anonymization and sanitization processes with automated machine learning-based synthetic data generation. This substitution reduces processing complexity by using learned patterns to automatically generate appropriate test data without manual intervention.
Data Source
AI summary
In some implementations, a testing system may receive a request for generation, based on a first dataset, of a second dataset, wherein the first dataset is associated with execution of a set of tests on an application, wherein the first dataset includes one or more data elements that satisfy one or more criteria for classification as private information. The testing system may process, using a machine learning model, the first dataset to identify one or more characteristics of the first dataset. The testing system may generate, using the machine learning model, the second dataset based on the first dataset, wherein the second dataset includes artificially generated data elements. The testing system may transmit an output identifying the second dataset.