How F5 Is Unlocking the Power of AI

At F5, we believe the future of applications will be multicloud, API-first, and powered by AI. At AppWorld 2024, we announced the F5 AI Data Fabric, a new data platform to deliver on the promise of AI. This innovation is a big step toward putting AI to work to deliver better app and API protection, more quickly, and with a ridiculously easy experience for users.

AI is obviously top of mind for everyone right now. Like any exciting technology, it can bring both benefit and risk. At F5, we’re innovating in a way that cuts to the heart of making AI both easy and safe to consume and use. With the F5 AI Data Fabric, we'll be accelerating the adoption of AI across our portfolio, and I believe some very specific applications of AI will change the game when it comes to security for apps and APIs.

Let’s start with how it works. The F5 AI Data Fabric offers a collection of services designed to help generate insights from massive amounts of data from products across our portfolio. From this data, we can deliver reports and analytics, or even train and deploy machine learning models to run inference to better secure and optimize apps. This approach provides the means to aggregate data from F5 Distributed Cloud Services, NGINX, and BIG-IP, surfacing insights via the Distributed Cloud Console.

A key component of this data fabric is the AI ecosystem, which is a collection of models, modules, and APIs designed to make machine learning and AI much more accessible to our product teams and users—ridiculously easy, as we say. It accelerates our ability to build AI into our products and go from idea to AI-enabled app super fast.

Here’s an example. If you’ve ever had to train supervised learning models, you know that it requires lots of data, and that data has to be labeled. Finding good, labeled data is a challenge, and labeling data properly can take weeks of toil. That’s not fast enough when you’re building models to keep up with a rapidly changing threat landscape.

Enter AI and large language models. In order to address this challenge, we take advantage of sentence transformers for vectorizing a training data set, and semantic search to bootstrap and automate the data labeling process. We can use similar techniques to then train and refine a model. Let’s say I want to build a classifier to detect SQL injection attacks. With just one example of an attack, the F5 AI Data Fabric can now help walk me through the process of labeling my data, training a model, and packaging it up in a container for deployment. What could previously take days or weeks now takes minutes (and I’m not even a data scientist!).

Once we've trained and deployed that classifier model, we can turn again to AI for help. Using large language models trained to create regular expressions, we can generate signatures that can be deployed out to the edge for real-time detection. Push-button signature generation allows us to respond very quickly as new threats emerge.

Leveraging these capabilities within our AI ecosystem allows us to dramatically shrink the time it takes to build a new detection model and then generate a signature from it. So, we can be super responsive to emerging threats. If there's a new attack type, a subject matter expert can quickly add detection capabilities into the data platform. Generative AI helps us deliver and accelerate better detection, at a lower cost to us and our customers.

And that’s just one of the many use cases AI can power for us.

Above, I described how the F5 AI Data Fabric can help us adapt and respond to threats more quickly. Generative AI can also dramatically improve the experience for our users.

To put this technology in the hands of customers, we’ll be introducing an AI assistant later this year that will give customers an intelligent partner to help accomplish their goals. Through a sophisticated natural language interface, users will be able to ask questions of their data and generate dashboards in real time. They’ll be able to generate product configurations, or even get help implementing a policy. Powered by the F5 AI Data Fabric, this tool will act as an embedded customer support manager, allowing customers to ask questions and receive recommendations based on the collective knowledge available to F5.

Users will be asking questions about their data like: “How many attacks did I see on my infrastructure last week? How many of those were SQL injection attacks? Where were they coming from?” It’s a far easier way for customers to interact with their data, get answers quickly, generate visualizations, and implement recommended remediations. A user will be able to ask questions like, “How do I configure bot protection for my new app?” and a chatbot will summarize how to do it. Beyond just guidance, this will also enable users to say things like, “Generate for me a configuration for a load balancer that has WAF enabled,” and it will perform the task.

In the coming months, we’ll begin exposing this functionality to Distributed Cloud Services customers, and I couldn’t be more thrilled to share what we’re building.