F5 has expanded its platform’s security capabilities to block exploitation attempts while a vulnerability still doesn’t have a definitive patch applied to the application. The company is adding anomaly detection, agentic threat intelligence, and new virtual patching features to its Web Application Firewall (WAF), aiming to shrink the gap between detecting a flaw and actually protecting the service.
The key facts about F5’s new virtual patching
- F5 is strengthening its WAF with anomaly detection, machine learning, and agentic threat intelligence.
- Virtual patching makes it possible to block specific exploitation paths without immediately modifying the vulnerable code.
- In internal testing, F5 says it achieved 98% detection effectiveness and a 1% false-positive rate.
- Distributed Cloud Web App Scanning can generate protections targeted at detected vulnerabilities.
- A virtual patch is a temporary measure: the underlying vulnerability still needs a definitive fix.
The approach responds to a problem that’s gaining weight as AI is adopted in cybersecurity. The same tools that help researchers and developers find bugs can also cut down the work needed to analyze a vulnerability, study its exploitation conditions, and develop attacks.
F5 argues that today’s most advanced AI models are shrinking the time between the discovery of a vulnerability and its exploitation. Facing that scenario, the company proposes shifting part of the response to the application’s own traffic: stopping a malicious request from arriving while developers prepare, test, and roll out the real patch.
What a Virtual Patch Is and What It Can Protect
A virtual patch doesn’t modify the vulnerable software.
Instead, it introduces a rule or security control in front of the application to recognize and block requests that try to exploit a given flaw. A WAF sits in a particularly good spot to do this, since it inspects requests before they reach the protected application.
This can be useful when an organization knows about a vulnerability but can’t immediately update the system.
A conventional fix can require modifying code, testing dependencies, checking that the new version doesn’t introduce regressions, and going through internal change-management procedures. In enterprise applications that need to stay available, that process can take longer than it takes an attacker to start probing a freshly disclosed vulnerability.
F5 wants to shrink that gap.
Its system lets teams use existing signatures or create custom rules tied to a CVE (Common Vulnerabilities and Exposures), an attack path, an HTTP method, a header, or a specific parameter.
The protection can be applied through F5 WAF for Distributed Cloud and extended to hybrid environments with F5 WAF for BIG-IP.
But there’s an important difference between mitigating an exploitation attempt and fixing a vulnerability.
If a WAF rule stops a malicious request from reaching the vulnerable code, it reduces exposure to that attack vector. The original bug is still present in the application and should be fixed with a definitive update whenever possible.
That’s why F5 frames virtual patching as a way to buy time, not as a permanent substitute for vulnerability management.
AI Analyzes Every Request and Looks for Anomalous Behavior
F5 is also strengthening the analysis it runs before deciding which traffic to block.
Its WAF uses machine-learning classification and a neural risk engine that evaluates requests as they arrive and assigns them a score.
The goal is to reduce sole reliance on known signatures.
Signatures remain useful when there’s an identifiable attack pattern, but they have limitations against unknown attempts or variations specifically designed to evade static rules.
F5 argues that dynamic scoring can help detect zero-day exploitation attempts, injection attacks, and polymorphic exploit chains, though actual effectiveness will depend on the type of attack, the application, and the configuration used.
The company is now adding anomaly detection to Distributed Cloud WAF.
This feature learns each application’s usual statistical patterns and compares new requests against that baseline. A deviation doesn’t automatically mean there’s an attack, but it can provide another signal for assessing a request’s risk.
This approach also introduces one of the common problems of any behavior-based detection system: telling malicious activity apart from legitimate but infrequent usage.
False positives are especially sensitive in a WAF. A tool that generates too many alerts can simply turn into noise. One that blocks legitimate traffic can directly affect customers and services.
F5 says that, in its own internal testing, its AI-powered WAF achieved 98% detection effectiveness and cut false positives to 1%. These are vendor-reported results and shouldn’t be interpreted as guaranteed levels for any given application or production environment.
From Finding a Vulnerability to Blocking Its Exploitation
Another piece of the strategy is connecting the discovery of a problem with the protection applied at runtime.
F5 Distributed Cloud Web App Scanning (WAS) can look for exposed vulnerabilities, unprotected APIs, and certain business-logic flaws. When it identifies an issue that can be mitigated, it can be used to trigger a specific virtual patch.
The idea is to shorten a process that traditionally requires several stages: a scanner discovers a vulnerability, generates an alert, the security team reviews it, determines its priority, and then configures a protective measure.
F5 aims to automate part of that path.
It’s adding to this agent-based threat intelligence, built on technology from Fletch, a company acquired by F5.
The system combines external information about emerging threats or vulnerabilities under active exploitation with the events observed in each organization’s applications.
This tries to address another common problem for security operations centers: knowing about a new vulnerability doesn’t necessarily mean it poses the same risk to every company.
Context matters.
A critical vulnerability in a component an organization doesn’t use carries a different priority than one affecting an internet-facing application that’s already seeing exploitation attempts.
F5 wants the system to help determine which threats are relevant to each environment and to propose mitigations that can become virtual patches.
Automation, however, also raises the importance of keeping controls over blocking decisions. An incorrect recommendation applied automatically to production traffic can cause availability problems, so cutting down false positives and being able to review mitigations become especially important.
AI Is Shrinking the Time Available to React
The backdrop to this announcement goes beyond a new WAF release.
Artificial intelligence is speeding up different parts of security work. It can review code, analyze changes, spot vulnerable patterns, and help build proofs of concept. These capabilities benefit defenders, but some can also be used to study targets and speed up attack development.
That forces a rethink of how long a vulnerability can stay exposed while the conventional patch is on its way.
F5 proposes temporarily separating the two needs: mitigate the attack in traffic first, and fix the underlying problem in the software afterward, while still going through the procedures needed to properly test the update.
The company rounds out this model with F5 Insight for Application Delivery and Security Platform (ADSP), designed to help operations teams with updates to F5’s own infrastructure.
The new AI-based WAF capabilities are available in Distributed Cloud within F5 ADSP. Virtual patching and the integration between Distributed Cloud WAS and F5 WAF for BIG-IP are also available, while anomaly detection and agent-based threat intelligence are rolling out gradually to Distributed Cloud WAF customers.
The practical value of this approach will depend on something especially sensitive in security: whether the temporary protection is precise enough to block the attack without becoming a new source of outages.
A virtual patch can shrink the exposure window. But a basic rule that automation doesn’t change still applies: blocking the path to a vulnerability isn’t the same as eliminating it.
Frequently Asked Questions
What is a virtual patch in cybersecurity?
It’s a protection applied in front of vulnerable software, typically through technologies such as a WAF, to block specific exploitation paths without immediately modifying the original code. It’s used as a mitigation until a definitive fix can be installed.
Does F5 use artificial intelligence to decide which traffic to block?
F5 uses machine learning, a risk-scoring engine, and anomaly detection to analyze requests. It’s also adding agent-based threat intelligence to contextualize vulnerabilities and recommend mitigations.
Does a virtual patch replace a security update?
No. It can temporarily reduce exposure to certain attacks, but the vulnerability remains present in the software. The application still needs to be updated or fixed to remove the underlying problem.
How effective is F5’s AI-powered WAF?
F5 says its internal testing achieved 98% detection effectiveness with a 1% false-positive rate. These are metrics provided by the company itself, and results may vary in real-world environments.
Source: f5
