Re: [OAUTH-WG] RAR 05 - Token response with sensitive data in draft-ietf-oauth-rar-05
Thank you Torsten, that's a good addition, it helps to have that clarified. BR Jacob Den mån 6 sep. 2021 kl 16:05 skrev Torsten Lodderstedt < tors...@lodderstedt.net>: > Hi Jacob, > > and here is the PR https://github.com/oauthstuff/draft-oauth-rar/pull/79 for > review. > > Thanks for the proposed text. I modified it a bit because I think the AS > should only omit data (not mask) and data can be provided even if > considered sensitive as long as there is a reasonable purpose and a legal > basis. > > best regards, > Torsten. > > Am 06.09.2021 um 09:52 schrieb Jacob Ideskog : > > Yes, privacy considerations could be more explicit about this. It should > probably explicitly mention the token response and the Client. > > I also suggest clarifying in 7 or 7.1 that the token response MAY be less > explicit or even different than the authorization details issued in the > tokens. > > This is not simple, since it may lead the Client to think it has more (or > less) access than it actually does, but since the intended audience of the > authorized data is the RS it should be ok. > A scenario I see is that the client requests Account access based on > pseudonyms or names of the accounts ("accounts" : ["foo", "bar"]) . The AS > replaces these with the actual account numbers ("accounts" : ["123-123", > "234-BCD"]) so the RS doesn't have > to deal with those translations. So: in the token response the pseudonyms > are still used, but in the issued token the explicit account values are > used. > > Suggestion for section 7: > "The AS MAY omit, mask or hide values in the authorization_details to the > Client in the Token Response if these are deemed sensitive and of no > intended use for the Client." > > Something in that direction would make it more clear that it is allowed to > do so, and that the Token Response doesn't prevent the issued token from > containing sensitive data. > > /Jacob > > > Den lör 4 sep. 2021 kl 11:41 skrev Torsten Lodderstedt < > tors...@lodderstedt.net>: > >> The AS intentionally shares the list of accounts in the mentioned example >> with the client. The assumption is the client asks for access to some >> accounts and the user decides which accounts to grant the client access to. >> This means the AS is authorized by the user to share this data. >> >> The privacy considerations section already has text about sharing data >> with resource servers. I suggest to add some text re data sharing with >> clients. >> >> Would that work for you? >> >> > Am 04.09.2021 um 03:12 schrieb Justin Richer : >> > >> > This is a fair point... The privacy and security considerations talk >> about this a bit as I recall, but likely need to in more depth and >> specificity. This is an intentional message channel to the client from the >> AS, but if the AS is blindly sending all information it might be saying >> more than it means to say to an entity that doesn't need that detail to >> function. Scopes have similar issues, but this structure adds more >> opportunities for mistakes just due to the possible increased complexity. >> > >> > -Justin >> > >> > From: OAuth [oauth-boun...@ietf.org] on behalf of Jacob Ideskog [ >> jacob.ides...@curity.io] >> > Sent: Friday, September 3, 2021 10:42 AM >> > To: oauth >> > Subject: [OAUTH-WG] RAR 05 - Token response with sensitive data in >> draft-ietf-oauth-rar-05 >> > >> > Hi all, >> > >> > I have a question about section 7.0 and 7.1 in draft-ietf-oauth-rar-05 >> that describes the token response. >> > >> > The authorization_details values could be sensitive in their nature. >> The example in section 7.1 highlights this nicely. The accounts array is >> empty when the client requests it, but is enriched by the AS and returned >> to the client in the token response. >> > >> > This means that the AS may leak potentially sensitive information to >> the client in a new place. Before this was only possible in the ID Token or >> UserInfo or if the AS returned a JWT as an access token which the client >> popped open (even though it shouldn't). >> > >> > I understand that the spec considers this an option for the AS to >> enrich or not. I think the enrichment is good and necessary, but with the >> side-effect of it ending up in the token response it becomes an issue. >> > >> > Is the token response a mirror of the authorization_details claim in >> the c
Re: [OAUTH-WG] RAR 05 - Token response with sensitive data in draft-ietf-oauth-rar-05
Hi Jacob, and here is the PR https://github.com/oauthstuff/draft-oauth-rar/pull/79 <https://github.com/oauthstuff/draft-oauth-rar/pull/79> for review. Thanks for the proposed text. I modified it a bit because I think the AS should only omit data (not mask) and data can be provided even if considered sensitive as long as there is a reasonable purpose and a legal basis. best regards, Torsten. > Am 06.09.2021 um 09:52 schrieb Jacob Ideskog : > > Yes, privacy considerations could be more explicit about this. It should > probably explicitly mention the token response and the Client. > > I also suggest clarifying in 7 or 7.1 that the token response MAY be less > explicit or even different than the authorization details issued in the > tokens. > > This is not simple, since it may lead the Client to think it has more (or > less) access than it actually does, but since the intended audience of the > authorized data is the RS it should be ok. > A scenario I see is that the client requests Account access based on > pseudonyms or names of the accounts ("accounts" : ["foo", "bar"]) . The AS > replaces these with the actual account numbers ("accounts" : ["123-123", > "234-BCD"]) so the RS doesn't have > to deal with those translations. So: in the token response the pseudonyms are > still used, but in the issued token the explicit account values are used. > > Suggestion for section 7: > "The AS MAY omit, mask or hide values in the authorization_details to the > Client in the Token Response if these are deemed sensitive and of no intended > use for the Client." > > Something in that direction would make it more clear that it is allowed to do > so, and that the Token Response doesn't prevent the issued token from > containing sensitive data. > > /Jacob > > > Den lör 4 sep. 2021 kl 11:41 skrev Torsten Lodderstedt > mailto:tors...@lodderstedt.net>>: > The AS intentionally shares the list of accounts in the mentioned example > with the client. The assumption is the client asks for access to some > accounts and the user decides which accounts to grant the client access to. > This means the AS is authorized by the user to share this data. > > The privacy considerations section already has text about sharing data with > resource servers. I suggest to add some text re data sharing with clients. > > Would that work for you? > > > Am 04.09.2021 um 03:12 schrieb Justin Richer > <mailto:jric...@mit.edu>>: > > > > This is a fair point... The privacy and security considerations talk about > > this a bit as I recall, but likely need to in more depth and specificity. > > This is an intentional message channel to the client from the AS, but if > > the AS is blindly sending all information it might be saying more than it > > means to say to an entity that doesn't need that detail to function. Scopes > > have similar issues, but this structure adds more opportunities for > > mistakes just due to the possible increased complexity. > > > > -Justin > > ________________ > > From: OAuth [oauth-boun...@ietf.org <mailto:oauth-boun...@ietf.org>] on > > behalf of Jacob Ideskog [jacob.ides...@curity.io > > <mailto:jacob.ides...@curity.io>] > > Sent: Friday, September 3, 2021 10:42 AM > > To: oauth > > Subject: [OAUTH-WG] RAR 05 - Token response with sensitive data in > > draft-ietf-oauth-rar-05 > > > > Hi all, > > > > I have a question about section 7.0 and 7.1 in draft-ietf-oauth-rar-05 that > > describes the token response. > > > > The authorization_details values could be sensitive in their nature. The > > example in section 7.1 highlights this nicely. The accounts array is empty > > when the client requests it, but is enriched by the AS and returned to the > > client in the token response. > > > > This means that the AS may leak potentially sensitive information to the > > client in a new place. Before this was only possible in the ID Token or > > UserInfo or if the AS returned a JWT as an access token which the client > > popped open (even though it shouldn't). > > > > I understand that the spec considers this an option for the AS to enrich or > > not. I think the enrichment is good and necessary, but with the side-effect > > of it ending up in the token response it becomes an issue. > > > > Is the token response a mirror of the authorization_details claim in the > > corresponding access token, or can it be a masked version? > > > > Perhaps the security co
Re: [OAUTH-WG] RAR 05 - Token response with sensitive data in draft-ietf-oauth-rar-05
Yes, privacy considerations could be more explicit about this. It should probably explicitly mention the token response and the Client. I also suggest clarifying in 7 or 7.1 that the token response MAY be less explicit or even different than the authorization details issued in the tokens. This is not simple, since it may lead the Client to think it has more (or less) access than it actually does, but since the intended audience of the authorized data is the RS it should be ok. A scenario I see is that the client requests Account access based on pseudonyms or names of the accounts ("accounts" : ["foo", "bar"]) . The AS replaces these with the actual account numbers ("accounts" : ["123-123", "234-BCD"]) so the RS doesn't have to deal with those translations. So: in the token response the pseudonyms are still used, but in the issued token the explicit account values are used. Suggestion for section 7: "The AS MAY omit, mask or hide values in the authorization_details to the Client in the Token Response if these are deemed sensitive and of no intended use for the Client." Something in that direction would make it more clear that it is allowed to do so, and that the Token Response doesn't prevent the issued token from containing sensitive data. /Jacob Den lör 4 sep. 2021 kl 11:41 skrev Torsten Lodderstedt < tors...@lodderstedt.net>: > The AS intentionally shares the list of accounts in the mentioned example > with the client. The assumption is the client asks for access to some > accounts and the user decides which accounts to grant the client access to. > This means the AS is authorized by the user to share this data. > > The privacy considerations section already has text about sharing data > with resource servers. I suggest to add some text re data sharing with > clients. > > Would that work for you? > > > Am 04.09.2021 um 03:12 schrieb Justin Richer : > > > > This is a fair point... The privacy and security considerations talk > about this a bit as I recall, but likely need to in more depth and > specificity. This is an intentional message channel to the client from the > AS, but if the AS is blindly sending all information it might be saying > more than it means to say to an entity that doesn't need that detail to > function. Scopes have similar issues, but this structure adds more > opportunities for mistakes just due to the possible increased complexity. > > > > -Justin > > ____ > > From: OAuth [oauth-boun...@ietf.org] on behalf of Jacob Ideskog [ > jacob.ides...@curity.io] > > Sent: Friday, September 3, 2021 10:42 AM > > To: oauth > > Subject: [OAUTH-WG] RAR 05 - Token response with sensitive data in > draft-ietf-oauth-rar-05 > > > > Hi all, > > > > I have a question about section 7.0 and 7.1 in draft-ietf-oauth-rar-05 > that describes the token response. > > > > The authorization_details values could be sensitive in their nature. The > example in section 7.1 highlights this nicely. The accounts array is empty > when the client requests it, but is enriched by the AS and returned to the > client in the token response. > > > > This means that the AS may leak potentially sensitive information to the > client in a new place. Before this was only possible in the ID Token or > UserInfo or if the AS returned a JWT as an access token which the client > popped open (even though it shouldn't). > > > > I understand that the spec considers this an option for the AS to enrich > or not. I think the enrichment is good and necessary, but with the > side-effect of it ending up in the token response it becomes an issue. > > > > Is the token response a mirror of the authorization_details claim in the > corresponding access token, or can it be a masked version? > > > > Perhaps the security considerations section should be updated with a > statement with regards to the fact that the client may see claim data only > intended for the RS? > > > > Regards > > Jacob Ideskog > > > > > > > > -- > > Jacob Ideskog > > CTO > > Curity AB > > --- > > Sankt Göransgatan 66, Stockholm, Sweden > > M: +46 70-2233664 > > j<mailto:ja...@twobo.com>a...@curity.io<mailto:a...@curity.io> > > curity.io< > https://www.google.com/url?q=http://curity.io=gmail-imap=163132276000=AOvVaw0O7NO5RiGVK6v1SxLCSz4k > > > > --- > > > > ___ > > OAuth mailing list > > OAuth@ietf.org > >
Re: [OAUTH-WG] RAR 05 - Token response with sensitive data in draft-ietf-oauth-rar-05
The AS intentionally shares the list of accounts in the mentioned example with the client. The assumption is the client asks for access to some accounts and the user decides which accounts to grant the client access to. This means the AS is authorized by the user to share this data. The privacy considerations section already has text about sharing data with resource servers. I suggest to add some text re data sharing with clients. Would that work for you? > Am 04.09.2021 um 03:12 schrieb Justin Richer : > > This is a fair point... The privacy and security considerations talk about > this a bit as I recall, but likely need to in more depth and specificity. > This is an intentional message channel to the client from the AS, but if the > AS is blindly sending all information it might be saying more than it means > to say to an entity that doesn't need that detail to function. Scopes have > similar issues, but this structure adds more opportunities for mistakes just > due to the possible increased complexity. > > -Justin > > From: OAuth [oauth-boun...@ietf.org] on behalf of Jacob Ideskog > [jacob.ides...@curity.io] > Sent: Friday, September 3, 2021 10:42 AM > To: oauth > Subject: [OAUTH-WG] RAR 05 - Token response with sensitive data in > draft-ietf-oauth-rar-05 > > Hi all, > > I have a question about section 7.0 and 7.1 in draft-ietf-oauth-rar-05 that > describes the token response. > > The authorization_details values could be sensitive in their nature. The > example in section 7.1 highlights this nicely. The accounts array is empty > when the client requests it, but is enriched by the AS and returned to the > client in the token response. > > This means that the AS may leak potentially sensitive information to the > client in a new place. Before this was only possible in the ID Token or > UserInfo or if the AS returned a JWT as an access token which the client > popped open (even though it shouldn't). > > I understand that the spec considers this an option for the AS to enrich or > not. I think the enrichment is good and necessary, but with the side-effect > of it ending up in the token response it becomes an issue. > > Is the token response a mirror of the authorization_details claim in the > corresponding access token, or can it be a masked version? > > Perhaps the security considerations section should be updated with a > statement with regards to the fact that the client may see claim data only > intended for the RS? > > Regards > Jacob Ideskog > > > > -- > Jacob Ideskog > CTO > Curity AB > --- > Sankt Göransgatan 66, Stockholm, Sweden > M: +46 70-2233664 > j<mailto:ja...@twobo.com>a...@curity.io<mailto:a...@curity.io> > curity.io<https://www.google.com/url?q=http://curity.io=gmail-imap=163132276000=AOvVaw0O7NO5RiGVK6v1SxLCSz4k> > --- > > ___ > OAuth mailing list > OAuth@ietf.org > https://www.google.com/url?q=https://www.ietf.org/mailman/listinfo/oauth=gmail-imap=163132276000=AOvVaw2Fa1GyOiE6a7mRCghwMI5J smime.p7s Description: S/MIME cryptographic signature ___ OAuth mailing list OAuth@ietf.org https://www.ietf.org/mailman/listinfo/oauth
Re: [OAUTH-WG] RAR 05 - Token response with sensitive data in draft-ietf-oauth-rar-05
This is a fair point... The privacy and security considerations talk about this a bit as I recall, but likely need to in more depth and specificity. This is an intentional message channel to the client from the AS, but if the AS is blindly sending all information it might be saying more than it means to say to an entity that doesn't need that detail to function. Scopes have similar issues, but this structure adds more opportunities for mistakes just due to the possible increased complexity. -Justin From: OAuth [oauth-boun...@ietf.org] on behalf of Jacob Ideskog [jacob.ides...@curity.io] Sent: Friday, September 3, 2021 10:42 AM To: oauth Subject: [OAUTH-WG] RAR 05 - Token response with sensitive data in draft-ietf-oauth-rar-05 Hi all, I have a question about section 7.0 and 7.1 in draft-ietf-oauth-rar-05 that describes the token response. The authorization_details values could be sensitive in their nature. The example in section 7.1 highlights this nicely. The accounts array is empty when the client requests it, but is enriched by the AS and returned to the client in the token response. This means that the AS may leak potentially sensitive information to the client in a new place. Before this was only possible in the ID Token or UserInfo or if the AS returned a JWT as an access token which the client popped open (even though it shouldn't). I understand that the spec considers this an option for the AS to enrich or not. I think the enrichment is good and necessary, but with the side-effect of it ending up in the token response it becomes an issue. Is the token response a mirror of the authorization_details claim in the corresponding access token, or can it be a masked version? Perhaps the security considerations section should be updated with a statement with regards to the fact that the client may see claim data only intended for the RS? Regards Jacob Ideskog -- Jacob Ideskog CTO Curity AB --- Sankt Göransgatan 66, Stockholm, Sweden M: +46 70-2233664 j<mailto:ja...@twobo.com>a...@curity.io<mailto:a...@curity.io> curity.io<http://curity.io> --- ___ OAuth mailing list OAuth@ietf.org https://www.ietf.org/mailman/listinfo/oauth
[OAUTH-WG] RAR 05 - Token response with sensitive data in draft-ietf-oauth-rar-05
Hi all, I have a question about section 7.0 and 7.1 in draft-ietf-oauth-rar-05 that describes the token response. The authorization_details values could be sensitive in their nature. The example in section 7.1 highlights this nicely. The accounts array is empty when the client requests it, but is enriched by the AS and returned to the client in the token response. This means that the AS may leak potentially sensitive information to the client in a new place. Before this was only possible in the ID Token or UserInfo or if the AS returned a JWT as an access token which the client popped open (even though it shouldn't). I understand that the spec considers this an option for the AS to enrich or not. I think the enrichment is good and necessary, but with the side-effect of it ending up in the token response it becomes an issue. Is the token response a mirror of the authorization_details claim in the corresponding access token, or can it be a masked version? Perhaps the security considerations section should be updated with a statement with regards to the fact that the client may see claim data only intended for the RS? Regards Jacob Ideskog -- Jacob Ideskog CTO Curity AB --- Sankt Göransgatan 66, Stockholm, Sweden M: +46 70-2233664 j a...@curity.io curity.io --- ___ OAuth mailing list OAuth@ietf.org https://www.ietf.org/mailman/listinfo/oauth