"In the age of artificial intelligence (AI), the key to corporations' security is speed. Hackers are rapidly expanding their reach at machine speed, while corporations are still defending at human speed and are exposed to threats. F5 is building a defense system that learns and detects AI-based attacks in real time so hackers cannot get through the front doors of AI systems."
Kunaciilan Nallappan, F5 vice president of marketing for Asia-Pacific, China and Japan (APCJ), said at the "F5 AppWorld Seoul 2026" press briefing on the 23rd at the Westin Josun Seoul Parnas in Samseong-dong, Seoul, "As hackers automate their attacks using AI agents, existing defense systems can no longer block the threats."
According to a survey F5 conducted of its customers, as attackers rapidly scaled up attacks using automation last year, web attacks rose 77% and bot activity increased 150%. Nallappan said, "Large language models (LLMs) are creating new attack surfaces, and preparing for the commercialization of quantum computers is no longer theoretical but has entered the stage of actual planning."
In particular, as a "hybrid multicloud" environment—where corporations use external cloud services such as Amazon Web Services (AWS), Microsoft Azure and Google Cloud alongside their own data centers—has become the standard, corporations face the need to secure consistent control capabilities across their distributed environments.
F5 redesigned its web application firewall (WAF) to respond to the new security threat landscape. Nallappan said, "Today's hackers create new payloads (malicious code) every second to launch attacks, but traditional WAFs were built on known threats and cannot respond," adding, "We added a new neural network–based layer to the WAF so the defense can stand up to attackers' real-time learning and attacks."
He added, "The neural network built on F5's own data can analyze threats across modern application environments in microseconds (one-millionth of a second), and because it operates entirely on a Neural Processing Unit (NPU) basis, it can even run at the network edge."
In AI systems, when a user enters a prompt (request), it goes into the application that serves as the door (front door), passes through orchestration (work coordination such as assigning AI agent roles), then performs inference and produces results.
F5 cited as a strength that its AI-based security capabilities provide strong security at three key control points: the "front door," the entry point to applications; "orchestration," where AI agents and data are connected; and "inference," where actual AI inference occurs. He said, "In the AI era, corporations' competitiveness will be determined not by whether they secure the latest AI models, but by how well they understand and control the core control points of AI systems."
In particular, F5 strengthened AI orchestration security based on an "AI red team," designed to identify sophisticated vulnerabilities, and "AI guardrails," which help prevent the leakage of sensitive data or unsafe data outputs. Nallappan said, "Based on this, we generate 10,000 new AI signatures every month." An AI signature is the knowledge accumulated by detecting and analyzing new types of attacks targeting AI models, such as prompt injection (injecting manipulated prompts to distort a model's algorithms) and jailbreaks (techniques to bypass safeguards).
He emphasized that F5's AI-based security capabilities also help increase corporations' productivity and reduce expense. Nallappan said, "The biggest constraint in today's artificial intelligence (AI) industry is energy," adding, "Today's AI factories are a kind of 'energy conversion machine' where tokens are output when energy is input, and unless we improve the energy efficiency of AI factories, corporations will waste electricity and profitability will inevitably deteriorate." Tokens are the basic units AI models use to process information and generate answers.
F5 improved energy efficiency by building a security platform that handles security and transmission tasks before answer generation at the data processing unit (DPU) stage, which comes before the graphics processing unit (GPU). He said, "An AI factory consists of CPUs for general computation, GPUs for inference and training, and DPUs for security and server distribution tasks," adding, "By processing security and transmission on the DPU before traffic reaches the GPU, F5 helped cut response generation time by 30% and increase token generation by 40%." By leveraging the DPU, the costly GPU can focus on inference work.
The company said it will strengthen its AI security platform capabilities by acquiring SurePass AI, which has technology to detect so-called "shadow AI" not approved by an organization.
Lee Hyeong-uk, head of F5 Korea, said, "The acceleration of AI adoption, the standardization of hybrid multicloud as an operating model, and the evolution of threats driven by the weaponization of AI are unfolding simultaneously," adding, "F5 will support Korean corporations in pursuing corporate innovation in a stable environment based on consistent security, performance and resilience."