Register to attend this P4 Developer Days webinar, “P4-based AI Feature Extractor for Volumetric Firewall Use Cases”
Date: July 9, 2025
Time 8:00am Pacific
Abstract
Using volumetry data is a very important feature of many firewalls from major security vendors. The changing security landscape and a need for performance and flexibility drive the use of AI/ML engines working more efficiently. The presentation shows an example of the use of P4 language to prepare a training feature and an example of an inference engine using P4 as a feature extractor for the SYN-FLOOD attack detector. The P4, as a packet processing language, can be used to define the network traffic-based features for AI/ML processing. The AI/ML engines can be abstracted as P4 externs. The P4 defines what packet fields are of interest in AI/ML process. Such P4 engine implementation (based on hardware or software) can also prepare/preprocess the AI/ML engine input (for example, rate of some packets or counter selection). It can also prepare the packet-based response using AI/ML external outcome. The presentation provides an overview of some challenges, like a lack of standardized datasets and performance requirements.
Speaker
Mirek Walukiewicz is Principal Engineer at Altera with more than 32 years of experience in networking technology. He is a Solution Architect actively working with customers on various P4 applications using FPGA-like security, packet scheduling, clock synchronization, and more. He drives P4 efforts inside Altera.