An Interactive Framework for Implementing Privacy-Preserving Federated Learning: Experiments on Large Language Models

Federated learning (FL) enhances privacy by keeping user data on local devices. However, emerging attacks have demonstrated that the updates shared by users during training can reveal significant information about their data. This has greatly thwart the adoption of FL methods for training robust AI models in sensitive applications. Differential Privacy (DP) is considered the gold standard for safeguarding user data. However, DP guarantees are highly conservative, providing worst-case privacy guarantees. This can result in overestimating privacy needs, which may compromise the model's accuracy. Additionally, interpretations of these privacy guarantees have proven to be challenging in different contexts. This is further exacerbated when other factors, such as the number of training iterations, data distribution, and specific application requirements, can add further complexity to this problem. In this work, we proposed a framework that integrates a human entity as a privacy practitioner to determine an optimal trade-off between the model's privacy and utility. Our framework is the first to address the variable memory requirement of existing DP methods in FL settings, where resource-limited devices (e.g., cell phones) can participate. To support such settings, we adopt a recent DP method with fixed memory usage to ensure scalable private FL. We evaluated our proposed framework by fine-tuning a BERT-based LLM model using the GLUE dataset (a common approach in literature), leveraging the new accountant, and employing diverse data partitioning strategies to mimic real-world conditions. As a result, we achieved stable memory usage, with an average accuracy reduction of 1.33% for $\epsilon = 10$ and 1.9% for $\epsilon = 6$, when compared to the state-of-the-art DP accountant which does not support fixed memory usage.

Privacy-Preserving and Trustworthy Deep Learning for Medical Imaging

The shift towards efficient and automated data analysis through Machine Learning (ML) has notably impacted healthcare systems, particularly Radiomics. Radiomics leverages ML to analyze medical images accurately and efficiently for precision medicine. Current methods rely on Deep Learning (DL) to improve performance and accuracy (Deep Radiomics). Given the sensitivity of medical images, ensuring privacy throughout the Deep Radiomics pipeline-from data generation and collection to model training and inference-is essential, especially when outsourced. Thus, Privacy-Enhancing Technologies (PETs) are crucial tools for Deep Radiomics. Previous studies and systematization efforts have either broadly overviewed PETs and their applications or mainly focused on subsets of PETs for ML algorithms. In Deep Radiomics, where efficiency, accuracy, and privacy are crucial, many PETs, while theoretically applicable, may not be practical without specialized optimizations or hybrid designs. Additionally, not all DL models are suitable for Radiomics. Consequently, there is a need for specialized studies that investigate and systematize the effective and practical integration of PETs into the Deep Radiomics pipeline. This work addresses this research gap by (1) classifying existing PETs, presenting practical hybrid PETS constructions, and a taxonomy illustrating their potential integration with the Deep Radiomics pipeline, with comparative analyses detailing assumptions, architectural suitability, and security, (2) Offering technical insights, describing potential challenges and means of combining PETs into the Deep Radiomics pipeline, including integration strategies, subtilities, and potential challenges, (3) Proposing potential research directions, identifying challenges, and suggesting solutions to enhance the PETs in Deep Radiomics.