Technical Name |
In-network Decision Tree for Intrusion detection in P4 Switches |
Project Operator |
National Taiwan University |
Project Host |
沈上翔 |
Summary |
A decision tree is implemented in a switch by using P4 language. Multiple forwarding tables are used to implement decision tree branch. Therefore, our solution can achieve high accurate malicious traffic detection with low network latency. |
Scientific Breakthrough |
Traditional switch can only detect malicious traffic by matching packet fields that limits the ability of the detection. We rely on existing switch to run decision tree in switches to detect malicious traffic with higher accuracy and lower latency. |
Industrial Applicability |
Switch providers can apply our solution to their product and add malicious traffic detection and prevention abilities in their switches. In addition, network service providers can leverage the switch with our solution to provide more accurate network traffic scanning without affecting network performance. The latency requirement of 5G networks can be achieved, while network security is provided. |
Keyword |
Machine Learning Decision Tree Network Security Forwarding Table P4 Language Intrusion Prevention Network Latency 5G network SDN NFV |