Hostname: page-component-7b7567c8c-7pwfn Total loading time: 0 Render date: 2026-11-15T08:56:24.291Z Has data issue: false hasContentIssue false

Hush hush: Keeping neural networks claims modelling private, secret, and distributed using federated learning: Discussion

Published online by Cambridge University Press:  25 June 2026

Rights & Permissions [Opens in a new window]

Abstract

Information

Type
Sessional Meeting Discussion
Creative Commons
Creative Common License - CCCreative Common License - BY
This is an Open Access article, distributed under the terms of the Creative Commons Attribution licence (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted re-use, distribution, and reproduction in any medium, provided the original work is properly cited.
Copyright
© The Institute and Faculty of Actuaries, 2026. Published by Cambridge University Press on behalf of The Institute and Faculty of Actuaries
Figure 0

Figure 1.Smartphone federated learning pipeline.

Figure 1

Figure 2.Insurance federated learning pipeline.

Figure 2

Figure 3.Need to encrypt parameters.

Figure 3

Figure 4.Application to reinsurance.

Figure 4

Figure 5.CMI application.

Figure 5

Figure 6.Data on French motor claims.

Figure 6

Figure 7.Insurance federated learning use case.

Figure 7

Figure 8.Neural networks.

Figure 8

Figure 9.Neural Network model set up.

Figure 9

Figure 10.Global Model Scenario – 10 insurers, 1 model.

Figure 10

Figure 11.Partial Model Scenario – 10 insurers, 1 model.

Figure 11

Figure 12.Federated Model Scenario – 10 insurers, 1 model.

Figure 12

Figure 13.Comparison of results.

Figure 13

Figure 14.Further comparison of results.

Figure 14

Figure 15.Comparison of insurer results.

Figure 15

Figure 16.Limitations.

Figure 16

Figure 17.What’s the catch?