Fork me on GitHub
Teleport

Reducing the Blast Radius of Attacks

Improve

Teleport encourages users to practice defense in depth so that every component of their infrastructure is safe against attacks, even if an attacker is partially successful. You can configure Teleport to add layers of protection to your cluster when users authenticate or request elevated privileges. In this guide, we will show you how to:

Make MFA mandatory for tsh login

If a user sets up an account to authenticate to their Teleport cluster with only a password, an adversary can gain access to the password using brute-force attacks, person-in-the-middle attacks, or phishing. But even if a user's password is compromised, you can stop an attacker from authenticating with it when they run tsh login.

Teleport lets you make it mandatory for a user to enroll an MFA device when they create an account, and to authenticate using that device when they begin a new Teleport session.

To do so, make the following changes depending on your environment:

Add the following to your Teleport configuration file:

auth_service:
  authentication:
    second_factor: otp|u2f|webauthn|on

Create the following cluster_auth_preference dynamic resource:

kind: cluster_auth_preference
version: v2
metadata:
  name: cluster-auth-preference
spec:
  second_factor: otp|u2f|webauthn|on

Create your dynamic resource using tctl create -f <path to your YAML file>.

To make MFA mandatory for all users, second_factor must be set to one of the following values:

  • otp
  • u2f
  • webauthn
  • on

Choose on if you would like to require MFA for all users while letting them choose an OTP, U2F, or WebAuthn device. The other options restrict users to a single type of MFA device, which is useful for enforcing a particular standard of security. Once you start the Teleport Proxy Service with the second_factor configuration option set to one of these values, Teleport will mandate MFA by:

  • Adjusting the Teleport signup page so a user must enroll an MFA device of the kind you have selected. If the value of second_factor is on, users will have the option to select from multiple device types.
  • Presenting the user with an MFA challenge when they run tsh login.

If you have enabled SSO in your Teleport environment, Teleport will conduct its own MFA challenge before sending a request to the identity provider. This means that, even if an attacker has managed to compromise your organization's SSO provider (e.g., by impersonating an administrator), a malicious user will still need to authenticate using valid credentials.

Warning

If your second_factor configuration is set to off and a user creates an account without a second factor, changing second_factor to a value that requires MFA will force that user to authenticate with a credential they have not registered. This will lock them out of their account. You have two ways to avoid this scenario:

  • Set second_factor to optional until you have confirmed that existing users have enrolled their MFA devices.
  • Run the tctl users reset <account> command to force a user to enter new credentials, including any required MFA device.

Present an MFA challenge for every attempt to access a resource

After a user logs into a Teleport cluster, they can request access to a particular resource, e.g., a node, database, application, or Kubernetes cluster. In this case, the Teleport Auth Service issues a single-use certificate for accessing that resource. You can prevent attackers from doing damage with a compromised certificate by enabling per-session MFA. With this setting, whenever a user requests a one-time certificate to access a resource, the Teleport Auth Service will issue an MFA challenge, even if the user has already begun a Teleport session via tsh login.

To enable per-session MFA for all users, do the following depending on your Teleport environment:

Make the following changes to your Teleport configuration file:

auth_service:
  authentication:
    require_session_mfa: yes

Create the following cluster_auth_preference dynamic resource:

kind: cluster_auth_preference
version: v2
metadata:
  name: cluster-auth-preference
spec:
  require_session_mfa: yes

Create your dynamic resource using tctl create -f <path to your YAML file>.

Require dual authorization for role requests

Even if an attacker gains access to a user's credentials and successfully signs into your Teleport cluster, you can still prevent the user from escalating their privileges. If you enable dual authorization, users who request to assume a particular role must obtain permission to do so from two or more reviewers. This way, if a malicous user manages to impersonate a legitimate one, reviewers can contact the real user before granting the new role.

Dual authorization uses Teleport's access plugins—e.g., Slack, JIRA, and PagerDuty—to notify reviewers that a user has requested a role. For access plugins that require a SAML or OIDC connector, you must enable the Cloud or Enterprise versions of Teleport.

You can set up dual authorization by applying two dynamic resources:

The reviewer

You can enable some users to review other users' role escalation requests by applying a dynamic resource similar to the following:

kind: role
version: v4
metadata:
  name: reviewer
spec:
  allow:
    review_requests:
      roles: ["role-one", "role-two", "role-three"]

Assign spec.allow.review_requests.roles to a list of role names. When a user requests access to one of the roles listed in spec.allow.review_requests.roles, your Teleport access plugins notify users with the reviewer role of the request and relay the responses to your Teleport cluster.

The reviewee

You can require a user to request access from reviewers by applying a dynamic resource similar to the following:

kind: role
version: v4
metadata:
  name: reviewee
spec:
  allow:
    request:
      roles: ["role-one", "role-two", "role-three"]
      thresholds:
        - approve: 2
          deny: 1

The spec.allow.request.roles field lists the names of other roles that a user with the reviewee role can request. When a reviewee requests access to one of these roles, Teleport notifies reviewers via your access plugins. The spec.allow.requests.roles.thresholds field indicates how many reviews are required to approve or deny the request.

Automatically prevent some roles from requesting others

A malicious Teleport user could request a more privileged role and trick a reviewer into granting access. You can prevent such a scenario by defining roles that prohibit users from even requesting access to particular roles.

The spec.deny field has the same possible properties as the spec.allow field we described earlier except, rather than enabling actions, this field disables them. For example, the spec.deny.requests.roles field is a list of roles that a user is prohibited from requesting. Teleport gives deny rules precedence over allow rules when executing access requests.

As an illustration, we have assigned user myuser to the user role, which we defined using the following template:

kind: role
version: v4
metadata:
  name: user
spec:
  deny:
    request:
      roles: ['admin']

Next, myuser attempts to request the admin role.

tsh request create --roles=admin

However, the Auth Service denies the request.

Creating request...

ERROR: user "myuser" can not request role "admin"

Restrict role requests based on user traits

Teleport's role resource lets you take precautions against accidental privilege escalation by ensuring that any user with particular attributes will have restricted access to certain roles. You can assign a list of traits to a user, then define a role resource that prevents any user whose traits match a regular expression from requesting elevated privileges.

A user has the same traits regardless of the roles they acquire. As a result, if a user happens to obtain another role because of an RBAC oversight, you can use trait-based restrictions to stop them from requesting a role with even more privileges.

Let's say that you have defined the following role for a contractor you have hired to analyze financial data.

kind: user
version: v2
metadata:
  name: myuser
spec:
  roles:
    - analyst # An unprivileged role
  traits:
    logins:
      - myuser
    groups:
      - contractors

Analysts sometimes need write access to your organization's database in order to create stored procedures, and can request access to the db-writer role. Only trusted analysts can request this access, and belong to a special admins group. Using deny rules, you can prevent analysts who are not in the admins group from requesting access to the db-writer role:

kind: role
version: v4
metadata:
  name: analyst
spec:
  deny:
    request:
      claims_to_roles:
        - claim: groups
          value: "{{regexp.not_match(\"admin\")}}"
          roles: ["db-writer"]
  allow:
    request:
      roles: ["db-writer"]
      thresholds:
        - approve: 2
          deny: 1

The claims_to_roles field within an allow or deny rule maps a user's traits to roles that they are either permitted or forbidden to request. In this case, we use the {{regexp.not_match(\"admin\")}} template function to prevent any user from requesting the db-writer role unless they have a groups trait with a value like administrator or admins. Users who do have such a trait can request the role with two approvals.

Set up your RBAC without admin roles

You can design your Teleport RBAC so that there is no all-powerful administrator in the system, or even a reviewer role with elevated privileges. This way, you can reduce the blast radius if an attacker successfully impersonates a Teleport user and requests a more privileged role.

First, define a role with privileged but limited access. In the following example, the editor role can log in as editor on hosts in our infrastructure in addition to the logins defined when creating the user. To prevent abuse, certificates issued to the user will be valid for only half a working day.

kind: role
version: v4
metadata:
  name: editor
spec:
  options:
    max_session_ttl: 4h
  allow:
    logins: [editor, "{{internal.logins}}"]

Next, we define the general user role. Users with this role can review other users' requests to become an editor, and can request the editor role themselves with two approvals. However, this user cannot log in as editor within our infrastructure.

kind: role
version: v4
metadata:
  name: user
spec:
  allow:
    logins: ["{{internal.logins}}"]
    review_requests:
      roles: ['editor']
    request:
      roles: ["editor"]
      thresholds:
        - approve: 2
          deny: 1
  deny:
    logins: ["editor"]

Two users can grant elevated privileges to another user temporarily without the need for a separate reviewer role that can become a target for attacks.

Next steps

Guides

Background reading

Have a suggestion or can’t find something?
IMPROVE THE DOCS