[Openid-specs-ab] Issue #1049: backchannel logout requests should include a reference to the OP (openid/connect)

Phil Hunt phil.hunt at oracle.com
Sat Sep 22 17:59:49 UTC 2018


It would be good if some people can talk this through face to face at IIW. Unfortunately I have to sit this one out as travel is on hold for the foreseeable future. Apologies. 

Phil

> On Sep 22, 2018, at 10:28 AM, Hans Zandbelt <hans.zandbelt at zmartzone.eu> wrote:
> 
> I believe that requiring mutual TLS would hinder adoption, certainly on the RP side since your average RP may not terminate TLS.  And also if they do it is typically just cumbersome or impossible to interwork with a different layer and configure client cert validation. Making it optional would be nice. The DoS risk is about the same as on other OAuth 2.0 endpoints IMHO.
> 
> I tend to agree that encryption may not be that useful and could be left out, some text should change, especially in:
> https://openid.net/specs/openid-connect-backchannel-1_0.html#Validation
> and also the overall suggestion that logout_token validation resembles id_token validation becomes a bit of a stretch considering the list of constraints then.
> 
> Hans.
> 
>> On Sat, Sep 22, 2018 at 6:33 PM Phil Hunt via Openid-specs-ab <openid-specs-ab at lists.openid.net> wrote:
>> Mutual tls (non tokenized) seems like a good solution provided the logout event node can use the OP’s issuer cert as client cert to establish the TLS connection. But this might not be easy for some microservice architectures as this means wider shared access to issuer private keys are needed which weakens overall security. 
>> 
>> Regarding encrypted events...
>> I don’t see the value in encrypting logout events. The logout event only contains identifiers that are transitory and are now end of life notices. The primary risk is an attacker serving fake logouts as part of a DoS attack. Signed logouts plus TLS transport should be enough here. 
>> 
>> We must also consider the general load on resource servers if they have to do a lot of crypto to reject false events. This is a DoS risk. Is Mutual TLS a good way to mitigate this or does it just shift the load to tls terminators? 
>> 
>> Phil
>> 
>>> On Sep 22, 2018, at 8:53 AM, Tom Jones <thomasclinganjones at gmail.com> wrote:
>>> 
>>> Wouldn't it make more sense for all back channel connex to be over tls with mutual auth?
>>> 
>>> thx ..Tom (mobile)
>>> 
>>>> On Fri, Sep 21, 2018, 8:38 AM Phil Hunt via Openid-specs-ab <openid-specs-ab at lists.openid.net> wrote:
>>>> Interesting. My assumption is iss, aud etc are req’d claims from JWT. 
>>>> 
>>>> However maybe a reminder is important?
>>>> 
>>>> Phil
>>>> 
>>>> > On Sep 21, 2018, at 4:52 AM, Hans Zandbelt via Openid-specs-ab <openid-specs-ab at lists.openid.net> wrote:
>>>> > 
>>>> > New issue 1049: backchannel logout requests should include a reference to the OP
>>>> > https://urldefense.proofpoint.com/v2/url?u=https-3A__bitbucket.org_openid_connect_issues_1049_backchannel-2Dlogout-2Drequests-2Dshould-2Dinclude&d=DwICAg&c=RoP1YumCXCgaWHvlZYR8PZh8Bv7qIrMUB65eapI_JnE&r=na5FVzBTWmanqWNy4DpctyXPpuYqPkAI1aLcLN4KZNA&m=lm9I-tIhoNwvye6UOWEMPW8NY3NHLUhJ9SotrZMkfjo&s=3LFWnJR17VF0dS5xSTSpiGzBJQ6AFN3Pu3Oa8M3ONMQ&e=
>>>> > 
>>>> > Hans Zandbelt:
>>>> > 
>>>> > Whilst taking a stab at implementing backchannel logout according to:
>>>> > https://urldefense.proofpoint.com/v2/url?u=https-3A__openid.net_specs_openid-2Dconnect-2Dbackchannel-2D1-5F0.html&d=DwICAg&c=RoP1YumCXCgaWHvlZYR8PZh8Bv7qIrMUB65eapI_JnE&r=na5FVzBTWmanqWNy4DpctyXPpuYqPkAI1aLcLN4KZNA&m=lm9I-tIhoNwvye6UOWEMPW8NY3NHLUhJ9SotrZMkfjo&s=L6lYaopVVDpj0Pk2gtvli2CrojHXip4pHWm-fsGlHyQ&e=
>>>> > 
>>>> > I found that for an RP that connects to multiple OPs it would be impossible to deduct the OP from the `logout_token` if it is encrypted with a symmetric key. Since following the OpenID Connect `id_token` guidelines (as suggested) it would have to decrypt with the `client_secret` which is (hopefully) a per-provider setting. Trying all OPs/`client_secret`'s consecutively would be very inefficient and probably not what anyone would want to do.
>>>> > 
>>>> > I suggest to add an `iss` parameter to the backchannel logout request in addition to the `logout_token` parameter. 
>>>> > 
>>>> > This will also make it easier for implementors to share the code path with `id_token` validation since they'd no longer have to "peek" into the `id_token` before calling the validation routine that may be issuer specific. The issuer would typically be known before validating the id_token since it is recorded in the (browser bound) state.
>>>> > 
>>>> > 
>>>> > _______________________________________________
>>>> > Openid-specs-ab mailing list
>>>> > Openid-specs-ab at lists.openid.net
>>>> > https://urldefense.proofpoint.com/v2/url?u=http-3A__lists.openid.net_mailman_listinfo_openid-2Dspecs-2Dab&d=DwICAg&c=RoP1YumCXCgaWHvlZYR8PZh8Bv7qIrMUB65eapI_JnE&r=na5FVzBTWmanqWNy4DpctyXPpuYqPkAI1aLcLN4KZNA&m=lm9I-tIhoNwvye6UOWEMPW8NY3NHLUhJ9SotrZMkfjo&s=QFeT_kOuXhKRo7gZWzW_kdBxaAC_PCO1A2u3BadpGqo&e=
>>>> 
>>>> _______________________________________________
>>>> Openid-specs-ab mailing list
>>>> Openid-specs-ab at lists.openid.net
>>>> http://lists.openid.net/mailman/listinfo/openid-specs-ab
>> _______________________________________________
>> Openid-specs-ab mailing list
>> Openid-specs-ab at lists.openid.net
>> http://lists.openid.net/mailman/listinfo/openid-specs-ab
> 
> 
> -- 
> hans.zandbelt at zmartzone.eu
> ZmartZone IAM - www.zmartzone.eu
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openid.net/pipermail/openid-specs-ab/attachments/20180922/e2823ca3/attachment-0001.html>


More information about the Openid-specs-ab mailing list