The Proposed Two-Stages of Zero-Knowledge-Proof-Based Anomaly Detectionby@quantification

The Proposed Two-Stages of Zero-Knowledge-Proof-Based Anomaly Detection

tldt arrow
Read on Terminal Reader
Read this story w/o Javascript

Too Long; Didn't Read

Cross-Round Check: The algorithm initializes with reference models, detecting potential attacks in Federated Learning rounds. It skips the check if no reference models exist. Cross-Client Anomaly Detection: Utilizing the three sigma rule, this stage assesses potentially malicious clients. L2 scores guide model removal, and an approximate average model is computed for subsequent rounds, ensuring robust security in Federated Learning.

Company Mentioned

Mention Thumbnail
featured image - The Proposed Two-Stages of Zero-Knowledge-Proof-Based Anomaly Detection
Quantification Theory Research Publication HackerNoon profile picture

@quantification

Quantification Theory Research Publication

The publication about the quantity of something. The theory about why that quantity is what is. And research!


Receive Stories from @quantification

react to story with heart
Quantification Theory Research Publication HackerNoon profile picture
by Quantification Theory Research Publication @quantification.The publication about the quantity of something. The theory about why that quantity is what is. And research!
Read my stories

RELATED STORIES

L O A D I N G
. . . comments & more!