Enhancing security & reliability of deployments in Kubernetes

Abstract

This project aims to further enhance the existing deployment infrastructure of NTU Speechlab’s automatic speech recognition (ASR) system in the following ways: reliability and security. The implemented solutions will enhance and provide security and reliability quality assurance to the system’s performance at any given point of time. We will be proposing three different types of testing, first of which will be load testing which aims to simulate varying levels of traffic while collecting the relevant metrics indicating how well the engine is performing under various loads. Next, we will be performing chaos testing with chaos mesh in order to verify the ASR system’s resilience in face of system failures. Lastly, we will be performing security testing based on the principles of DevSecOps which not only cultivates the element of security at every stage of the Software Development Lifecycle but allows for secure code to be written and deployed in a shorter period of time. For load and security testing, we will be automating it via Tekton pipelines whereas for chaos testing, we will be demonstrating how to perform a full chaos experiment. This report will include full details and implementation of the solutions described along with the relevant architectural diagrams. The report will also include an analysis of any experiment carried out and the implication it has on the ASR system.