Transformer and GAN Detection for Sensitive Data Leakage
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Microservice architectures in enterprise computing systems face challenges with sensitive data leakage, leading to high false positive rates and latency due to inefficient data detection methods like regular expressions, which lack nuance and precision.
Innovation Solution
Implementing generative adversarial networks (GANs) and transformer models to detect and prevent sensitive data leakage by applying a blocking policy, reducing false positives and enhancing system performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If regular expressions are used to detect sensitive data in microservice responses, then data leakage can be prevented, but false positive rates increase and detection precision decreases
Solution Approach 1:
The patent replaces traditional regular expression-based detection (mechanical pattern matching) with machine learning models including BERT transformers and GANs. These models learn semantic patterns and contextual relationships in data, enabling more accurate distinction between sensitive and non-sensitive information while reducing false positives.
Solution Approach 2:
The patent transforms the detection approach by changing from fixed regex patterns to adaptive machine learning parameters. The GAN framework allows the system to learn optimal detection parameters through adversarial training, improving precision while maintaining reliability in preventing actual data leakage.
2Productivity
If traditional detection methods are used across hundreds of microservice endpoints, then system coverage is maintained, but processing latency increases
Solution Approach 1:
The patent implements GAN-based pre-training where the generator and discriminator are trained beforehand on extensive datasets. This preliminary action creates optimized detection models that can be rapidly deployed across multiple microservice endpoints without requiring real-time complex computations, thereby reducing latency while maintaining comprehensive coverage.
3Reliability
If regular expression patterns are applied to all data traffic, then potential sensitive data can be identified, but false positive rates become unmanageably high
Solution Approach 1:
The patent substitutes rigid regular expression matching with flexible machine learning-based detection. The BERT transformer models understand contextual semantics and can distinguish between legitimate uses of sensitive-looking patterns (false positives) and actual sensitive data, dramatically reducing false positive rates while maintaining identification reliability.
Data Source
AI summary
Disclosed herein are system, method, and computer program product embodiments for increasing data security by using generative adversarial networks (GAN) and transformer models to detect sensitive data leakage. A transformer model may receive a message via a network. The transformer model may then apply a GAN model to determine whether the message contains potentially sensitive data requiring further inspection. If the message contains potentially sensitive data, the transformer model may determine sensitive data types may be included in the message and the likelihood that sensitive data is present. Based on the transformer model's determination, a blocking policy may be applied to prevent sensitive data leakage.