Information Flow Control in Machine Learning through Modular Model Architecture
This addresses the challenge of secure machine learning for access-controlled data, enabling training on sensitive information while maintaining strict control over data influence, which is incremental in applying IFC principles to ML architectures.
The paper tackles the problem of uncontrolled information flow in machine learning models when training on sensitive data with access control, by proposing an information flow control (IFC) definition and a modular Transformer architecture that limits data influence to specific expert modules. The result shows minimal performance overhead (1.9%) and significant accuracy improvements (up to 62%) on text and code datasets.
In today's machine learning (ML) models, any part of the training data can affect the model output. This lack of control for information flow from training data to model output is a major obstacle in training models on sensitive data when access control only allows individual users to access a subset of data. To enable secure machine learning for access-controlled data, we propose the notion of information flow control for machine learning, and develop an extension to the Transformer language model architecture that strictly adheres to the IFC definition we propose. Our architecture controls information flow by limiting the influence of training data from each security domain to a single expert module, and only enables a subset of experts at inference time based on the access control policy.The evaluation using large text and code datasets show that our proposed parametric IFC architecture has minimal (1.9%) performance overhead and can significantly improve model accuracy (by 38% for the text dataset, and between 44%--62% for the code datasets) by enabling training on access-controlled data.