Move from Detect to Prevent in K8s with many Ingress Rules

Larger K8s environments can have many ingress rules specified within the ingress resource(s) or routes defined for Kong gateway.

Here's a suggested workflow explaining how to set open-appsec to prevent to protect them after initial onboarding was done in detect mode and sufficient time has passed to gain confidence about open-appsec and allow the solution to perform the learning.

Start with your own detect policy, see following example: (Or create a new one and apply to your ingress, see Configuration using CRDs)

apiVersion: openappsec.io/v1beta1
kind: Policy
metadata:
  name: example-policy
  # BEFORE moving certain ingress rules to prevent-learn
spec:
  default:
    triggers:
    - appsec-special-log-trigger
    mode: detect-learn
    practices:
    - webapp-best-practice
    source-identifiers: my-source-identifiers
    trusted-sources: my-trusted-sources
    custom-response: appsec-web-user-response-example
    exceptions:
    - appsec-exception-example

Once sufficient confidence as well as training is reached for certain resources to move them to prevent: Create specific rules in "prevent-learn" mode in the policy object for those resources (see also the hint below the following example). The learning recommendations as shown in the WebUI will provide helpful suggestions about when to move to prevent for each protected asset, see Track Learning and Move from Learn/Detect to Prevent In the following example the following hosts/path will be protected in "prevent-learn" mode, assuming ingress rules are defined as well to allow the traffic: - web.server.com/example - web.server.com/another-example - another.webserver.com The default mode for all other resources based on existing ingress rules will remain "detect-learn", as specified in the default section of the policy example.

apiVersion: openappsec.io/v1beta1
kind: Policy
metadata:
  name: example-policy
  # AFTER moving certain ingress rules to prevent-learn
spec:
  default:
    triggers:
    - appsec-special-log-trigger
    mode: detect-learn
    practices:
    - webapp-best-practice
    source-identifiers: my-source-identifiers
    trusted-sources: my-trusted-sources
    custom-response: appsec-web-user-response-example
    exceptions:
    - appsec-exception-example
    
  specific-rules:
  - host: web.server.com/example
    triggers:
    - appsec-special-log-trigger
    mode: prevent-learn
    practices:
    - webapp-best-practice
    source-identifiers: my-source-identifiers
    trusted-sources: my-trusted-sources
    custom-response: appsec-web-user-response-example
    exceptions:
    - appsec-exception-example
  - host: web.server.com/another-example
    triggers:
    - appsec-special-log-trigger
    mode: prevent-learn
    practices:
    - webapp-best-practice
    source-identifiers: my-source-identifiers
    trusted-sources: my-trusted-sources
    custom-response: appsec-web-user-response-example
    exceptions:
    - appsec-exception-example
  - host: another.webserver.com
    triggers:
    - appsec-special-log-trigger
    mode: prevent-learn
    practices:
    - webapp-best-practice
    source-identifiers: my-source-identifiers
    trusted-sources: my-trusted-sources
    custom-response: appsec-web-user-response-example
    exceptions:
    - appsec-exception-example

It is possible to create kind of "super rules" in the open-appsec policy that match multiple specific rules in your ingress and reduces effort and complexity, as well of the size of the policy object. Example: For protecting multiple ingress rules that might exist in an ingress resource which are all for the same host "another.webserver.com" (but having different paths defined) it is sufficient to have a single host entry for "another.webserver.com" in the specific-rules section of the open-appsec policy object configured for "prevent-learn" mode (as shown in the above example). Those "super rules" are defined as any other rule in the specific-rules section. With this concept the amount of specific policy rules for open-appsec required to protect a complex environment in K8s can be much lower than the amount of ingress rules defined for it.