Accelerating Intrusion Detection Dataset Analysis- A Framework Using AutoGen Agents for CIC-IDS 2017

Abstract

An IDS is a vital component in securing any network, however, the practical operation of an IDC is often dependent upon reasonable response times for the data with a huge volume. In this paper, we attempt to enhance the analysis of the CIC-IDS 2017 dataset using AutoGen, a deep learning model framework related to state-of-the-art. AutoGen performs a lot of the work automatically without requiring human intervention bottlenecks such as data preprocessing, feature engineering, or even model training thus saving a lot of time and work when developing an IDS. We compared the performance of AutoGen against prompt-based language models by focusing on task completion metrics along with three additional metrics: Humane Evaluation score, time taken, and resource overhead. The results exhibited that AutoGen is far superior to conventional ones in every way possible. In summary, the findings of this study demonstrate AutoGen’s popularity for the future of intrusion detection through its data analysis function in the bias of the entire system performance parameter.