Mr Mitchell Hargreaves1, Mr Suhag Byaravalli Arun1, Mr Steve Quenette2
1Monash Data Futres Institute, Clayton, Australia, 2Monash eResearch Centre, Clayton, Australia
Through the use of Data Processing Units (DPUs), traditional processing of network traffic can be offloaded and processed in real time, returning computer processing power back to the user whilst increasing cybersecurity analysis. Using NVIDIA’s Morpheus, network traffic can be analysed with AI models, augmenting traditional reactive approaches to cybersecurity with adaptive Natural Language Processing and Forrest Inferencing. Moreover, Morpheus enables an organisation to maintain sovereignty over the cybersecurity algorithms applied to their systems. For example, we can expect further scrutiny of risk-vs-performance in HPC and IoT environments, tending towards more security. We can also expect more diverse workloads and the ability for research application-specific cybersecurity (tending towards a Zero-Trust security model).
While Morpheus’s off the shelf models boast high accuracies on their test sets, they need fine tuning for deployment. We deploy Morpheus onto the Nectar Research Cloud in a test environment, and make our work available as an image in Glance. We then show that this image can be used to fine tune NVIDIA’s Sensitive Information Detection model to detect potential information leaks with minimal effort. The image includes detailed documentation of how to reproduce the environment along with links to relevant documentation to each of the services in the Morpheus stack. This image will give students and researchers a working Morpheus environment out of the box, allowing them to participate in bringing real time and adaptable AI security pipelines to the world.
Biography:
Mitchell Hargreaves
Mitchell is a deep learning engineer with an interest in deep learning models, applied AI and making these tools more available for all. His experience includes developing deep learning solutions to research problems as well as teaching workshops on practical AI.
Suhag Byaravalli Arun
Suhag is a research DevOps specialist with an interest in MLOps. He previously worked on various aspects of Data Science/ML problems from fast-prototyping, to productionizing and maintaining solutions, primarily specialising in chipping away the “Hidden Technical Debt in Machine Learning”
Steve Quenette
Dr Steve Quenette is the Deputy Director of the Monash eResearch Centre. This multi-disciplinary centre now includes over 40 eResearch and IT professionals providing expertise, computing, visualisation and data capabilities into numerous research areas such as Cryo-electron microscopy, Macromolecular Crystallography, Neuroscience, Archaeology, Proteomics, Genomics, Structural Biology, Bio-medical Imaging, Climate Modeling, Computational Chemistry, Materials Engineering, Fluid Dynamics.