Note
Access to this page requires authorization. You can try signing in or changing directories.
Access to this page requires authorization. You can try changing directories.
Advanced Application Threat Protection analyzes eligible HTTP requests to help detect exploit attempts against web applications and APIs. It adds AI-based request analysis to Azure Firewall Premium's existing protections.
This article shows how to configure the capability through a security profile group and associate the group with application rules. Start in Alert only mode before using Alert and deny. For the shared model, see Azure Firewall AI Security concepts.
Important
Advanced Application Threat Protection is currently in PREVIEW. See the Supplemental Terms of Use for Microsoft Azure Previews for legal terms that apply to Azure features that are in beta, preview, or otherwise not yet released into general availability.
Prerequisites
- Azure Firewall Premium in a region where the AI Security preview is available. See Azure Firewall AI Security concepts.
- The AI Security add-on enabled on all firewalls associated with the firewall policy.
- An Azure Firewall Policy with application rules covering the intended HTTP or HTTPS traffic.
- For HTTPS, working TLS inspection and client trust in the inspection certificate authority. See Azure Firewall Premium certificates.
- An authorized test environment and a test application that you own, with approved test requests and a benign control.
Enabling the add-on on an existing firewall requires downtime. Complete that maintenance activity before starting this procedure; creating a profile group isn't a substitute for enabling the add-on.
During the preview, configuration is available only in the Azure portal.
Inspection scope
The capability inspects eligible HTTP or HTTPS requests that match application rules associated with the protection group. The described scenarios include east-west traffic between Azure workloads and traffic from on-premises networks to Azure workloads, routed through Azure Firewall.
Traffic translated by inbound DNAT rules isn't processed by application rules and isn't inspected by this capability.
Application-request analysis and file analysis address different content. For supported file transfers, see Configure Advanced Malware Threat Protection. Application threat protection complements rather than replaces Azure Web Application Firewall, application security testing, and timely patching.
Create a security profile group
Use the guided group experience to configure application threat protection.
- In the Azure portal, open the firewall policy.
- Select Settings > Security Profile Groups (preview) > Create security profile group.
- On the Basics tab, enter a configuration name and select Advanced application threat protection (preview). Select Next.
- On the Configure protections tab, enter a profile name, or turn on Use existing Advanced application threat protection security profile and select a profile.
- For Mode, select Alert only for the initial evaluation. The other options are Disabled and Alert and deny.
- Under Risk threshold, set the Alert percentage for the requests that you want to review.
- Select Next, review the settings, and then select Create.
An action is triggered when the request's risk score is greater than its configured threshold. Lower values trigger actions at lower scores. Choose thresholds based on your evaluation rather than treating an illustrative value as a recommended default.
Other capabilities in the group are optional. A group can contain at most one profile of each type.
Apply the group to application rules
Before changing a rule, check whether a matching network rule prevents the test traffic from reaching the intended application rule. Azure Firewall processes network rules first. For more information, see Azure Firewall rule processing.
- Select Rules > Application rules, and create or edit the target rule.
- Scope its sources and destinations to the intended evaluation traffic.
- Select the security profile group that you created.
- Configure the protocol:
- For HTTPS, use HTTPS:443 and enable TLS inspection on the rule.
- For HTTP, use HTTP:80. TLS inspection isn't required for unencrypted HTTP.
- Save the rule.
Verify inspection in Alert only mode
Evaluate the configured protection against a test application before enabling denial.
- Confirm routing through the firewall and a match on the intended application rule.
- Check the selected group, protection mode, and alert threshold.
- For HTTPS, confirm successful decryption and client trust in the inspection certificate authority.
- Send a benign control request and an approved application-security test request to the test application that you own.
- Review the recorded risk scores and actions for the requests. Compare them with the configured threshold.
No particular test request is guaranteed to produce a detection. A successful request alone doesn't prove inspection occurred. Existing Azure Firewall monitoring helps establish the traffic path, but validation also needs evidence of the application threat protection verdict.
Enable denial after evaluation
After reviewing the results in Alert only mode:
- Change the application's protection mode in the group to Alert and deny.
- Set both the Alert and Deny thresholds.
- Repeat the approved tests in the authorized environment.
- Confirm denial for a request whose risk score exceeds the deny threshold. A denied request receives HTTP 403 Forbidden and doesn't reach the application.
- Repeat the benign control and review the recorded actions before expanding the rule's scope.
If inspection doesn't produce the expected result
Check the following items before changing thresholds:
- The traffic traverses the firewall and matches an application rule, rather than a network or DNAT rule.
- The rule references the intended security profile group.
- The protection mode and thresholds match the evaluation.
- HTTPS traffic is successfully decrypted and the client trusts the inspection certificate authority.
Risk scores are probabilistic and can produce false positives or false negatives. Don't treat a low score or an absent alert as proof that a request is safe.