Version: v0.4.1 Latest Latest

This package is not in the latest version of its module.

Go to latest
Published: Jan 16, 2023 License: Apache-2.0

README reference Go Report Card Releases LICENSE Slack Status codecov


This readme file focus on assisting Knative users to secure their deployed services. When working with vanilla Kubernetes, you may deploy Guard by adding a sidecar.

Why Do We Need Security-Guard?

User containers deployed on Kubernetes may include vulnerabilities, may be misconfigured and may include malicious code. The source of such vulnerabilities, misconfigurations or malicious code may be the DevOps team, a dependency or a hacker that has successfully penetrated any of the support systems (image repository, ci/cd, Knative, Kube, the DevOps team development systems etc. etc. etc.) or one of the backend services used by the user container. Any such security issue may enable an attacker to use the user container for other purposes than its original intention (e.g. steal data, attack others, spread, contact a C&C, Crypto mining, etc.)

Users of Knative, same as any Kubernetes user, require the means to block (and/or get an alert about) an attempt to exploit a vulnerability or misconfiguration, embedded in a user container. Also, users require the means to establish situational awareness about container running potentially malicious code and be offered ways to respond once they discover that the user containers are being exploited by attackers.

How Does Security-Guard Help Secure Knative Services

The core component of Security-Guard, or Guard for short, is the guard-gate which can be used to extend Knative Queue Proxy. guard-gate monitors and potentially blocks requests and/or responses to Knative services based on a per-service security configuration.

The per service security configuration is stored in a "Guardian" object. Guardian maintains a set of micro-rules that enable a fine grain filtering, performed against each value delivered to/from the service. By using the micro-rules, guard-gate can identify the delivery of exploits targeting vulnerabilities embedded as part of the service or its dependencies.

When guard-gate and a proper set of micro-rules are used, offenders will typically be required to build dedicated delivery mechanisms in order to explore options for detecting and exploiting service vulnerabilities. This lengthy process may need to be repeated for each service, as each service maintains a different set of micro-rules. As a result, an offender will not be able to use common statistical attack patterns.

Users of guard-gate gain Situational Awareness both thanks to alerts about out-of-pattern request/responses, and by the identification of security indicators that the service is misused. Such indicators include longer than usual service times and the list of external IP addresses approached by the service.

Beyond, Situational Awareness, guard-gate enables blocking out-of-pattern behaviors and the ability to react to potential attacks and/or to on-going attacks by introducing a fine-tune configurable security gate in front of the service.

Overall, the solution offers both visibility into the security of the service, and the ability to monitor/block both known patterns and unknown patterns (using a zero-day exploits).

Solution Components

This project adds:

  1. guard-gate - the workload security gate that is located in front of each service
  2. A Guardian - an object implemented as either a CRD or a Configmap maintaining a set of micro-rules per service
  3. guard-service - a learner service to auto learn the micro-rules stored in the Guardian
  4. guard-ui - a user interface web app to simplify manual configuration of micro-rules

In addition, the project adds:

  1. qpoption - a package to enable using guard-gate with Knative Queue Proxy Option
  2. test-gate - a package to enable testing Knative Queue Proxy Options named
  3. pluginterfaces - a base package to enable registration and interfacing security gates

See Guard Architecture to learn about how Guard process and learn internally security data.

Guard Gate

guard-gate can be loaded as a knative queue proxy option using qpoption

Once loaded, it monitors the proxied requests and responses and the pod network.


guard-gate uses Guardian - a set of micro-rules that define the expected behavior of the service.

The Guardian object maintains the following configuration parameters:

  1. A control object which defines the working mode of Guard in relation to the service
  2. A learned set of micro-rules updated by guard-service
  3. A manually configured set of micro-rules that can be updated using kubectl or using guard-ui

Each service has its own respective Guardian. New services may use a default Guardian for the Kubernetes namespace or may use the default Guard Guardian. The following procedure describes how the Guardian for service servicename in namespace namespace is found. Note that guard-gate may be configured to either work with CRDs ( or with Configmaps guardian-* as Guardians.

When looking for a service Guardian:

  • A Dedicated-Guardian per service is used (CRD named <servicename> or Configmap named guardian-<servicename>).
  • If the Dedicated-Guardian is not found, a Namespace-Default -Guardian is used as a starting point (CRD named ns-<namespace> or Configmap named guardian-ns-<namespace>).
  • If a Namespace-Default-Guardian is not found, a Default-Guardian is used as a starting point.
  • Note that the Default-Guardian uses auto-learning to help bootstrap new Guardians.

See guard-gate for more details on the different Guard working modes.

Guard Service

guard-service is a standalone service used to learn Guardian micro-rules based on inputs from instances of guard-gate. guard-service stores the Guardian as a CRD ( under the name <servicename> or in a Configmap under the name guardian-<servicename>.

Additionally, guard-service cache Guardians and serve guard-gate requests to get a copy of the Guardian. This is done to reduce the load on KubeApi since any deployed Knative pod requires to access a copy of the Guardian on startup.

Guard User Interface

Although Guardian CRDs and Configmaps can be controlled directly via kubectl, an optional guard-ui is offered to simplify and clarify the micro-rules.


Guard takes a zero-trust approach. Guard assumes that all services are most likely vulnerable and places a gate in front of every service, implementing a zero-trust architecture.


Path Synopsis
Package v1alpha1 is the v1alpha1 version of the API.
Package v1alpha1 is the v1alpha1 version of the API.
This package has the automatically generated clientset.
This package has the automatically generated clientset.
This package has the automatically generated fake clientset.
This package has the automatically generated fake clientset.
This package contains the scheme of the automatically generated clientset.
This package contains the scheme of the automatically generated clientset.
This package has the automatically generated typed clients.
This package has the automatically generated typed clients.
Package fake has the automatically generated clients.
Package fake has the automatically generated clients.
Package lru provides three different LRU caches of varying sophistication.
Package lru provides three different LRU caches of varying sophistication.

Jump to

Keyboard shortcuts

? : This menu
/ : Search site
f or F : Jump to
y or Y : Canonical URL