switch-ml

Run traditional machine learning models for traffic classification on Tofino switches using P4. Reproduced and transplanted from the paper: Do Switches Dream of Machine Learning?: Toward In-Network Classification. Course project for undergraduate course Computer Network, 2025 fall at DCST, Tsinghua University.

Currently supports:

• Decision Tree Classifier
• K-Means Clustering
• Naive Bayes Classifier
• Support Vector Machine (SVM) Classifier

For the first 2 models, the data plane codes in P4 are at ./switch, and the control plane codes in C are at ./ctrl.

For Naive Bayes and SVM, their codes are at ./bayes and ./svm respectively. Please refer to the README files in those directories for more details.

Compile and Run

Compile the code using root:

make
./decision_tree # or ./kmeans

This compiles both the data plane and control plane program, and runs the control plane program for the Decision Tree Classifier. You can replace decision_tree with kmeans to run the K-Means Clustering controller program.

For Naive Bayes and SVM, please refer to their respective README files for compilation and execution instructions.

The controller program will:

• Load the P4 program to the Tofino switch data plane
• Enable the necessary ports
• Insert rules into the switch tables