K3Y’s federated intrusion detection and Prevention System
K3FL1DPS is a federated solution for detecting and discriminating various threats and cyberattacks. Based on network traffic data and system logs, K3FL1DPS incorporates multiple deep intrusion/anomaly detection models into a federated level.
K3FL1DPS platform offers the following services
– Anomaly Detection Based on Secure Knowledge Sharing
The DL-based models are able to discriminate a wide range of cyberattacks, anomalies and malware, relying on network flow statistics, application-layer features and binary data, thus representing local models that subsequently will be incorporated into a federated architectural schema.
– Personalised Solution
It is noteworthy that the federated model is personalised in each use case, taking full advantage of active learning.
– Centrally orchestrated Federated Learning
K3FL1DPS follows a centralised federated learning approach composed of (a) local Deep Learning (DL)-based models, (b) federated clients and (c) federated server.
– Adaptive Federated Learning Architecture
The Federated Server is in charge of: (a) orchestrating the federated training procedure, (b) querying the Federated Clients and finally, and c) retrieving, aggregating and distributing the federated models. The Federated Client is responsible for training the local models under the directions of the Federated Servers.