Summary of Vflgan-ts: Vertical Federated Learning-based Generative Adversarial Networks For Publication Of Vertically Partitioned Time-series Data, by Xun Yuan and Zilong Zhao and Prosanta Gope and Biplab Sikdar
VFLGAN-TS: Vertical Federated Learning-based Generative Adversarial Networks for Publication of Vertically Partitioned Time-Series Data
by Xun Yuan, Zilong Zhao, Prosanta Gope, Biplab Sikdar
First submitted to arxiv on: 5 Sep 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Cryptography and Security (cs.CR)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary In the realm of artificial intelligence (AI), dataset quality is paramount for training high-performing models. However, data sharing constraints due to privacy concerns and regulations hinder model development. A potential solution involves releasing synthetic datasets with similar distributions to private datasets. Nevertheless, scenarios exist where attributes required for AI model training are distributed among parties that cannot share local data for synthetic data construction due to privacy regulations. The authors of this article recently introduced the Vertical Federated Learning-based Generative Adversarial Network (VFLGAN) for publishing vertically partitioned static data in PETS 2024. However, VFLGAN is ineffective in handling time-series data, which presents both temporal and attribute dimensions. To address this limitation, the authors propose VFLGAN-TS, a combination of attribute discriminators and vertical federated learning to generate synthetic time-series data in vertically partitioned scenarios. The performance of VFLGAN-TS approaches that of its centralized counterpart, representing the upper limit for VFLGAN-TS. To ensure privacy protection, the authors apply Gaussian mechanisms to satisfy (,)-differential privacy and develop an enhanced privacy auditing scheme to evaluate potential privacy breaches through the framework of VFLGAN-TS and synthetic datasets. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Imagine you want to train a super smart AI model, but you can’t share your data with others because it’s private. This is a big problem in artificial intelligence. One solution is to create fake data that looks like the real thing. But what if different people have pieces of information they need to share, and they can’t share them because of privacy rules? The authors of this article came up with a new way to solve this problem using something called VFLGAN-TS. This method helps make fake time-series data, which is like making fake weather forecasts or stock market predictions. It’s really good at doing this job! To keep the data private, they use special tricks to make sure it stays safe. They also created a way to check if their method works well and doesn’t put people’s privacy at risk. |
Keywords
» Artificial intelligence » Federated learning » Generative adversarial network » Synthetic data » Time series
