IBM Research – Finance Track
11todo: 1test..Abstract Instructions Please fill in the sections below with information about your proposed solution. Then save the document as a PDF and include your name or your team name in the file name of the saved document. Please upload your completed document in the PETs Prize Challenge: Phase 1 abstract submission form. Note that complete abstracts should adhere to the following style guidelines:
- Abstracts must not exceed one page in length. - Abstracts may be submitted in PDF format using 11-point font for the main text. - Set the page size to 8.5”x 11” with margins of 1 inch all around. - Lines should be at a minimum single-spaced.
Solution Description A brief description of your proposed solution, including the proposed privacy mechanisms and architecture of the federated model. The abstract should describe the proposed privacy-preserving federated learning model and expected results with regard to accuracy. Successful abstracts will outline how solutions will achieve privacy while minimizing loss of accuracy.
Team
| Name | ID | Name | ID |
|---|---|---|---|
| Alan King | kingaj12 | Your Name | here |
| Nathalie Baracaldo | kingaj12 | Your Name | here |
| Nir Drucker | nir.drucker | Your Name | here |
| Hayim Shaul | hayim | Your Name | here |
Solution Description
The finance track of the PETs Prize Competition focuses on the detection of payment anomalies. Anomalies can have many causes. Some causes are mundane: perhaps the beneficiary account has been closed. Other causes are more serious: perhaps the owner of the account is flagged as suspicious. The entities in this challenge are of two types. One type is a payment provider (SWIFT) which has access to the payment details: the ordering and beneficiary accounts, routing information, and payment anomaly labels. The other type (Banks) have confidential information (Flags) about the accounts or account owners. The goal of the challenge is to develop a privacy-preserving transaction anomaly prediction model, possibly also indicating the probable cause. Our initial assessment shows that the confidential Account Flags are at the top of the feature importance list. Therefore, a significant part of our solution design will be directed to a Vertical Federated Learning solution that can safely consume Flags data.
In a few paragraphs, please write a brief description of your proposed solution. The abstract should:
-
•
describe the proposed privacy mechanisms
-
•
describe the architecture of the privacy-preserving federated learning model
-
•
outline how solutions will achieve privacy and security while minimizing loss of accuracy
-
•
defend the use of novel techniques