Experimental Systems for Studying Network Attacks and Their Application in Machine Learning

The paper discusses experimental systems developed for studying network attacks and their use in machine learning models training. An analysis compares the effectiveness of an experimental network in a virtual environment versus a real network. The results demonstrate improved data collection speed and the capability of isolated traffic generation for various network threats, enhancing attack classification accuracy. The use of tools such as Kali Linux, Wireshark and iperf3 proved cost-effective and highly adaptable. Findings are supported by comparative experiments and suggestions for method improvement. This study highlights the value of isolated traffic for training neural networks and other machine learning models, concluding with an assessment of the proposed systems' potential.