Improve protocol inference rules

[oazizi000]

Pixie’s protocol tracer captures traffic for various protocols, including HTTP, gRPC, postgres and many more. The protocol tracer relies on protocol inference rules to classify traffic based on the contents of the messages being sent.

The protocol inference rules sometimes misclassify traffic, which results in observability gaps.

To validate the accuracy of our protocol inference rules, we need a large database of different traffic patterns (we call this dataset TrafficNet). For this task, we need to (1) expand the TrafficNet data set to include more samples of existing and new protocols, and (2) experiment with different inference models to improve the accuracy of the protocol inference.

[mory91]

Hi @oazizi000 !. I would be happy to work on this project. Can you provide some details and resources to look for, for a better proposal?

[PrashantSaxena000000]

Hello! I would like to contribute to this project for GSOC'22. Can you please help me to get started with it?😊

[htroisi]

@mory91 @PrashantSaxena000000 - It's great to hear that you're interested in working on Pixie! As a next step, could you join Pixie's Community Slack and DM Omid to discuss further?