This article is brought to you by NVIDIA
Last week, 10 cyber development teams gathered – seven from Israel and the rest from the UK and India – with one goal: to win the NVIDIA cyber hackathon. They had 24 hours at their disposal, during which they had to develop innovative security applications based on the three core technologies developed by the company. Some of the winning developments may be incorporated into future NVIDIA products.
Some of the teams came from academia and some from the industry, and they were accepted to the hackathon after presenting innovative ideas in the cyber world based on the company's technologies. During the event and before it (on Bootcamp day), the participants received tools, training, and examples of using the various technologies, and were closely accompanied by NVIDIA’s engineers and product experts.
The first-place team won an NVIDIA Bluefield-2 card and $2,000. Second and third place won $1,500 and $800, respectively. In addition, for the first time, a Young Entrepreneur award was given for $500.
The developments had to be based on the BlueField DPU data processing chip, which accelerates communication, storage, and cybersecurity tasks in data centers while offloading the workloads from the traditional processor (CPU), or the graphics processor (GPU); and also NVIDIA DOCA – BlueField's software architecture, which provides APIs, drivers, libraries, sample code, documentation, services, and containers, and allows to accelerate the development and implementation of DPU applications in data centers. And since we’re talking about it, here is a friendly reminder that the DOCA SDK is currently available for early access here.
The third technology that the teams had to build on is Morpheus – an open software platform for developing security applications, which allows cyber developers to create artificial intelligence-based processes for filtering, processing, and classifying large amounts of data in real-time. It uses AI processing on GPUs to detect and act against cyber threats or security anomalies.
The Young Entrepreneur Award
The award was given to team 8200-2B – high school students who take part in the technological leadership organization "After-Tech". They worked on a cyber security solution to detect malicious log-ins during the attack – unlike most solutions that detect the hack only after it has occurred. The team used Morpheus and GPU acceleration for data filtering, processing, and classification at scale.
Third place went to the Ariel-2 team from Ariel University, which worked on a solution to detect malware in encrypted traffic, based on the Morpheus platform and acceleration performed by a graphics processor (GPU). Using Deep Learning, the team created a training model that allows data to be classified as suspected malware even if it is encrypted and demonstrated effective machine learning and AI training at reduced costs. The uniqueness of the model is the selective use of the characteristics of the network traffic and not necessarily of the packet itself.
Second place went to the GAPU team from Octopus ComputerSolution and the security sector, which developed a new layer of security management on top of the DPU, between the server and the network infrastructure. The development creates a first line of defence against malware, based on the DOCA architecture and can be modularly adjusted and implemented on a large scale as needed. The solution is called ARMadillo, after the ARM cores in the BlueField chip, and the hard armour of the unusual animal. It uses DOCA FLOW to speed up security processes and illustrates how security tasks can be offloaded from the CPU and host memory directly to the DPU. The layer of protection provided by ARMadillo includes a "5-Tuple" firewall, a DNS filter, and deep packet inspection.
The big winners are the members of the "Yahlom" team from the C4I Corps, who managed to create an advanced load balancer based on DOCA FLOW APIs within 24 hours. The development combines the advantages of a hardware-based load balancer with a software-based solution that uses the NVIDIA BlueField DPU to accelerate large-scale network traffic. The Load Balancer developed by the team includes support for dynamically adding or removing Nodes, as well as User-Defined load balancing. Using a Bluefield DPU to implement a load balancer illustrates the advantage of shifting network services outside the CPU and expanding the service with new capabilities without burdening the computing resources. With the DPU, the team created an additional compute unit external to the CPU that handles and accelerates network load management.