Mustafa Topsakal

The project provides a series of progressively designed laboratory experiments covering Ethernet, IP, ARP, DNS, UDP, TCP, and secure communication protocols. Students analyze real network traffic using Wireshark and command-line tools to observe protocol behavior and network dynamics. The laboratory is built on a layered, observation-driven methodology, where each experiment focuses on a specific network layer while reinforcing cross-layer interactions.

The work has two stages. First, normal and attack scenarios are run in a CoRE4INET-based OMNeT++ simulation environment to generate the dataset. Second, the framework converts raw simulation traces into labeled datasets, trains and evaluates six machine learning models with cross-validation, and optionally applies tSMOTE or tsaug augmentation to address class imbalance.

The workspace contains three scenarios: an in-vehicle network of cameras, multimedia and control units (31 streams), a larger in-vehicle network with radar, lidar and actuators (17 streams, 8 switches), and an industrial automation network with gPTP time synchronization (8 streams). The scenarios use Credit-Based Shaper traffic shaping for AVB Class A/B. The large in-vehicle network was ported from CoRE4INET to INET; the version used to reproduce the paper's results (OMNeT++ 6.3.0 + INET 4.5.4) is kept on a separate branch.