Realization of an Intrusion Detection use-case in ONAP with Acumos

With Software-Defined Networking and Machine Learning/Artificial Intelligence (ML/AI) reaching new paradigms in their corresponding fields, both academia and industry have exhibited interests in discovering unique aspects of intelligent and autonomous communication networks. Transforming such intentions and interests to reality involves software development and deployment, which has its own story of significant evolution. There has been a notable shift in the strategies and approaches to software development. Today, the divergence of tools and technologies as per demand is so substantial that adapting a software application from one environment to another could involve tedious redesign and redevelopment. This implies enormous effort in migrating existing applications and research works to a modern industrial setup. Additionally, the struggles with sustainability maintenance of such applications could be painful. Concerning ML/AI, the capabilities to train, deploy, retrain, and re-deploy AI models as quickly as possible will be crucial for AI-driven network systems. An end-to-end workflow using unified open-source frameworks is the need of the hour to facilitate the integration of ML/AI models into the modern software-driven virtualized communication networks. Hence, in our paper, we present such a prototype by demonstrating the journey of a sample SVM classifier from being a python script to be deployed as a micro-service using ONAP and Acumos. While illustrating various features of Acumos and ONAP, this paper intends to make readers familiar with an end-to-end workflow taking advantage of the integration of both open-source platforms.