Digital Twin and an Approach to its Change Management using Generative Modeling

Abstract

With the rise of cybersecurity threats, patches/changes to physical systems (i.e., computers) have become compulsory to make the system up-to-date and prevent any attack against it. However, these patches/changes cannot be directly applied to a production environment where critical services like security tools run. Larger business entities may have a separate production and non-production environment, but smaller business entities such as startups may not have such benefits. In this, research, we introduce a digital twin for IoT devices for smaller business entities. This digital twin acts as a non-production environment where patches/changes can be applied, and automated/manual tests can be performed before and after the production system is patched. Stress testing the digital twin after the patch is applied requires additional data to create functional tests or compare the behavior of the digital twin before the patch application with the behavior after the patch application. For this purpose, we designed a Network Activity ontology to guide a Generative Adversarial Network (GAN) with domain knowledge to generate synthetic network traffic data that could be used to identify test case scenarios to create functional tests manually. Additionally, only GAN was used to yield synthetic system data for future use in stress testing of digital twins for system-level defects after patch application. We designed basic functional tests using Snort rules and Scapy.