The world we live in has always had its challenges. Today, thanks in part to the data explosion as we are now firmly in the fourth, data-driven paradigm, the challenges are more complex, including how to protect and secure important data in this digital world.
From your own personal financial data to Personal Health Information (PHI) and more, protecting yourself personally has taken on an entirely new look. Likewise, in industry, competition is fierce, it is global, and data now plays a central role in every large business entity. Companies now invest heavily in security, which now means more than locks and guards — it means cybersecurity.
This article will give you some perspective by taking a brief look at what cybersecurity was and now is, providing an example of cybersecurity in industry, an overview of the infrastructure driving cybersecurity, and touch on what tomorrow could bring in terms of advancements.
Then and Now
Liz Joyce, CISO at HPE, did a great job of detailing what was industrial cybersecurity versus what it looks like now in How private industry views the cyberthreat problem (enterprise.nxt, 10 August 2018). “In earlier years, cyberprotection was straightforward. ‘Basically, we could take our critical assets, drop them in a data center, stick up a firewall, and feel good about things.’ Now, however, ‘we're in a highly connected mobile hybrid environment, and your data is sitting in a data center, a cloud, on devices on the edge,’ she said. ‘And for those devices, we've hit a tipping point... where the number of devices now significantly outnumbers the number of human beings on this planet. It's about 11 billion to about 7 billion, and all of those things have data. How are we going to protect that data?’” Great question!
The article continued as Joyce suggested that cybersecurity measures should be integrated into regular business practices. ”That means creating applications with security built in from the start. It means building security directly into firmware and hardware. It means using technologies like artificial intelligence and machine learning, and training every employee in a company about cyberthreats and how to avoid them. And ultimately, it means building security into the core of the enterprise itself.”
Using AI and ML alone begs of the need for HPC, due to quantity of data collected and needing analysis and protection, in managing cybersecurity, particularly in large organizations.
One Example of an Industrial Cybersecurity Implementation
Use Cases for Artificial Intelligence in High-Performance Computing (Tractica-Cray, 2Q 2017) details several possible cybersecurity solutions sets. “The success of deep learning in image classification and speech recognition has been extended to the cybersecurity space where algorithms learn about threats in real time and are especially good at detecting first-time malware, or changing attack surfaces without the need for laborious feature engineering to take place. Graph analysis has also shown a lot of promise and is well suited to represent cybersecurity data, which can be mapped by nodes, and is especially good at detecting botnet threats, port scans, or lateral movement in networks. However, graph analytics is highly compute-intensive and incapable of doing real- time analytics on a CPU cluster when running in large organizations, which can see as many as 1 to 2 billion events per day.
What Drives Today’s Cybersecurity
Traditionally, non-HPC cybersecurity has lived in IT organizations. It used to be about installing antivirus software throughout organizations. How far we have come, for better or for worse.
As complexity (i.e., data growth and tools to manage and protect it) has risen, so has the need for HPC to provide security. HPC cybersecurity has been managed predominantly with CPUs on the processor side but again, with the data deluge, we have seen a significant rise in GPU processing being a part of cybersecurity.
The next five years will be… interesting, if not incredibly challenging in the world of cybersecurity. In the Use Cases article, the need for ML/DL and, hence, HPC is clear: “Cybersecurity solution vendors that use machine learning and deep learning to process threats expect to see a rapid increase in the number of malicious software samples being used to train algorithms.” The article continues that modern workloads entail “a few million samples of malicious code, but the samples are expected to increase by 30X to 50X in the next 5 years. At that scale, using an enterprise-grade server would fall short. Using GPU-based HPC solutions will need to be adopted.”
Oh, the options before us all in protecting our important data resources going forward. It is clearly a moving target when it comes to how best to provide security. Many think that even the largest organizations or even governments are exposed to potential breaches. Let’s all hope that cybersecurity advancements can stay ahead of this varying game.