ML Workspace Privacy Profiling for Automated PII Masking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems face challenges in efficiently and automatically classifying and masking Personal Identifiable Information (PII) in machine learning (ML) workspaces, leading to increased data breach risks due to manual data classification, tedious annotation processes, and evolving compliance requirements.
Innovation Solution
An automated Preserving Privacy Engine (PPE) framework that integrates with ML workspaces to detect PII using deep neural networks and reinforcement learning, classify data based on compliance rules, and mask sensitive information, ensuring continuous compliance and security.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If manual data classification and annotation are used, then data privacy and compliance can be maintained, but productivity decreases and time consumption increases
Solution Approach 1:
The system enables self-service through automated PII detection and classification where the machine learning model autonomously identifies and classifies personal identifiable information without requiring manual intervention from data practitioners, thereby maintaining compliance while improving efficiency
Solution Approach 2:
The patent replaces manual mechanical data classification processes with an automated machine learning-based system that uses deep neural networks to detect, classify, and mask PII, eliminating the need for human annotators while maintaining accuracy and compliance
2Productivity
If automated PII detection is implemented, then productivity improves, but device complexity increases
Solution Approach 1:
The system segments the PPE framework into distinct functional modules including PII detection module, classification module, and masking module, each handled by specialized machine learning models. This segmentation allows independent development, testing, and optimization of each component while managing overall system complexity
Solution Approach 2:
The machine learning models are designed with multi-functionality to handle various PII types and compliance requirements within a unified framework. The same detection infrastructure can classify different data types and apply appropriate masking strategies, reducing the need for separate specialized systems
3Measurement precision
If deep neural networks are used for PII detection, then measurement precision improves, but use of energy increases
Solution Approach 1:
The system applies partial action by selectively deploying deep neural networks only for data that requires thorough analysis, while using lighter-weight algorithms for straightforward cases. This staged approach maintains high precision for complex PII detection while reducing overall computational energy consumption
Data Source
AI summary
One example method includes automatically scanning, at a privacy data collector, received data to determine if the received data is related to an Artificial Intelligence (AI)/Machine Learning (ML) workspace that is used to build an ML model. For the received data that is determined to be related to the AI/ML workspace, parsing the data, by the privacy data collector, to determine if the data includes any Personal Identifiable Information (PII) or other sensitive information. For the data that includes PII data or other sensitive data, generating, by a ML classification model, a privacy classification for the data. For the classified data, performing, by a data masking component, a data masking operation on the PII data or other sensitive data to generate masked data.