Skip to content

Latest commit

 

History

History
101 lines (71 loc) · 3.56 KB

0066-ip-based-validation.md

File metadata and controls

101 lines (71 loc) · 3.56 KB
authors state
Przemko Robakowski([email protected])
implemented (Teleport 10.0.0)

RFD 66 - IP-based validation

Note: this approach was refined in RFD 110

Required Approvers

  • Engineering @zmb3 && (@codingllama || @nklaassen )
  • Product: (@xinding33 || @klizhentas )

What

IP-based validation is an enterprise-only feature that embeds the source IP in SSH certificates. When enabled, SSH certificates can only be used to connect to resources from the same source IP that was issued the certificate.

Why

It provides additional security against leaked credentials, if adversary gets hold of certificate he won't be able to use them outside machine that created it. It also forms part of user identity.

Relevant issue: #7081

Details

Configuration

New field will be added to role options definition:

  • pin_source_ip - defines if certificate should be pinned to the IP of the client requesting it. User won't be able to use the certificate from different IP for example if one wants to move certificate to different machine without doing login there or if one uses mobile internet that changes IP addresses frequently or network with multiple exit nodes with different IPs.

Example configuration:

kind: role
metadata:
  name: dev
spec:
  options:
    pin_source_ip: true

Implementation

Following definition will be added to types.proto:

message RoleOptions {
  // ...

  // PinSourceIP defines if certificate should be pinned to the IP of the client requesting it.
  bool PinSourceIP = 18 [(gogoproto.jsontag) = "pin_source_ip", (gogoproto.casttype) = "Bool"];
}

If any role has PinSourceIP set to true then IP of the client requesting certificate will be encoded depending on certificate type:

  • SSH certificate will encode IP using source-address critical option as defined by OpenSSH. This option is recognized by sshd from OpenSSH and also by Go's ssh package, so it will be enforced automatically in Teleport.
  • TLS certificates (used by DB, Kubernetes, Application and Desktop access) already encode IP in custom extension with OID 1.3.9999.1.9 in tls/ca.go. It is then decoded as part of tlsca.Identity and will be validated in *authorizer.Authorize method lib/auth/permissions.go

Encoding above will happen in all places we generate certificates:

  • lib/auth/auth.go#generateUserCert
  • lib/auth/join.go#generateCerts (Machine ID)
  • lib/auth/auth_with_roles.go#generateUserCerts() (renewals, impersonation etc)

Implementation must ensure that all calls *authorizer.Authorize provide valid client IP (there are at least HTTP, gRPC and databases protocols to handle).

UX

This change should be mostly transparent for the user. Administrator will add relevant option to role definitions and should work for all users.

If user tries to use certificate on other machine (different IP) tsh will force relogin as it currently does when certificate expires:

~$ ./teleport/tsh ssh node1.cluster.local
Enter password for Teleport user admin:

Security

This proposal does not protect against IP spoofing, but it should provide at least the same level of security as we have today (as this additional protection, not replacement for user authentication). Likewise, it won't prevent attacks that use certificates from original machine as malware, worms etc.