We do something similar except we use an existing OAuth flow and simply add custom attributes to the authorization token. That authorization token is then sent along with requests to various services and these attributes are picked out and then used to apply policies or output filtering as appropriate.
As a suggestion I would look to name the properties of your current token in such a way where they could be compatible with the embedded case.
Instead of suggesting paseto, auth0 blog clearly says how to mitigate the vulnerability by using kid and specifically mentioning which algorithm during validation. I think most of the implementation do this way.
I built something similar for Neon[1]. We went with the route that the user sessions for JWT RLS had very few privileges, and that the context was managed by our ingress/connection pool. In our case, all incoming queries had to come via our sql-over-http interface, but that was mostly just to reduce scope rather than being integral to the security.
The general idea is that the connection pool ingress is trusted to provide the context properly, but the client is not. The pool would generate a random JWK for each connection and each request would use a newly signed JWT and the token ID is strictly monotonic, which makes any attempts for the client to smuggle their own token contexts in near impossible.
The JWTs the pool signs for each transaction will contain the claims that were correctly validated from the client.
A `DISCARD ALL` before returning the connection to the available pool will clear the context so it cannot be re-used and a new context must be provided fresh.
I did something similar about 20 years ago, although not using JWT.
The application supplied a session token when first querying the database. The database would make a call to the authentication server to verify it and obtain the roles associated with that session. If it had not seen that combination of roles before, it would build and cache access tables for them (needed for performance when using views for row level security). Then any further queries using that token were fast. This also worked with connection pooling.
It had some rough edges; we had to run periodic cleanup jobs to remove stale cached access tables and tokens. And even with them, enforcing row level security consumed about 30% of the database server performance!
But, it did work and remained in service for about 7 years. Eventually, they replaced it with something much simpler. I had left and nobody understood it anymore. Which I guess is a separate issue about doing unusual things with tech...
7 comments
[ 2.7 ms ] story [ 32.3 ms ] threadAs a suggestion I would look to name the properties of your current token in such a way where they could be compatible with the embedded case.
https://github.com/tvondra/jwt_context/blob/10be23c0651f1099...
https://github.com/tvondra/jwt_context/blob/10be23c0651f1099...
Oh look, the typical setup for a classical JWT vulnerability.
Prior art:
https://auth0.com/blog/critical-vulnerabilities-in-json-web-...
https://github.com/firebase/php-jwt/issues/351
You should really consider not using JWT for new designs that don't a priori need to interop with JWT.
PASETO is less likely to create sadness: https://paseto.io
The correct fix is to never trust the header for which algorithm you're using.
https://infosec.exchange/@sophieschmieg/115132001021790785
The general idea is that the connection pool ingress is trusted to provide the context properly, but the client is not. The pool would generate a random JWK for each connection and each request would use a newly signed JWT and the token ID is strictly monotonic, which makes any attempts for the client to smuggle their own token contexts in near impossible.
The JWTs the pool signs for each transaction will contain the claims that were correctly validated from the client.
A `DISCARD ALL` before returning the connection to the available pool will clear the context so it cannot be re-used and a new context must be provided fresh.
[1] https://github.com/neondatabase/pg_session_jwt/
The application supplied a session token when first querying the database. The database would make a call to the authentication server to verify it and obtain the roles associated with that session. If it had not seen that combination of roles before, it would build and cache access tables for them (needed for performance when using views for row level security). Then any further queries using that token were fast. This also worked with connection pooling.
It had some rough edges; we had to run periodic cleanup jobs to remove stale cached access tables and tokens. And even with them, enforcing row level security consumed about 30% of the database server performance!
But, it did work and remained in service for about 7 years. Eventually, they replaced it with something much simpler. I had left and nobody understood it anymore. Which I guess is a separate issue about doing unusual things with tech...