Background I am building a web app that allows the user to integrate with multiple services like Google, Twitter, Github etc. using OAuth2.0.

Currently, I retrieve the refresh token on sign-in to service and store it in my DB. (encrypted before storing). (Access Tokens are discarded after use).

The application is hosted on AWS, although the number of services available on AWS is overwhelming I have gone through them and selected KMS for encrypting the tokens in the app before writing to DB. I explored Secrets Manager but the pricing is not suitable to store user tokens.

Question I would like to know if there is a better way to store Refresh Tokens other than Encrypting and storing in a DB. If not, are there specific things to keep in mind to ensure security while using a DB for refresh tokens?


  • Flask Application
  • PostgreSQL DB (AWS RDS)

1 Answer 1


If I may, I'll abstract your question into "I have something short and secure that needs to be stored - where can I store it?" Secret Manager and RDS are both legitimate places to store small things. Other places to store stuff include: the onboard hard drive, Dynamo Db, and S3 buckets.

You mention that the tokens are discarded pretty quickly - the onboard hard drive might be a fine place to store this ephemeral data (but it doesn't scale out to multiple uses very well!)

Dynamo Db is VERY quick and can handle lots of different AWS services writing and reading at once - if you have a situation where a single user is getting access tokens to a dozen different services within seconds, concurrency is going to be a concern and Dynamo Db (also RDS) will handle this well. If you have a situation where LOTS of microservices hit APIs and rely on the access token, scaling is going to be a concern, and Dynamo Db will handle this well. MAKE SURE YOU DELETE THE OLD Access Tokens after a time.

S3 can handle scaling out very well - but you will need to architect for concurrency. That is, you should have a specific bucket and "folder" (S3 is key value, folders in S3 aren't real) for the Access Tokens, and have a separate file for each user. A single file will contain each of the Access Tokens that user has logged into - and then on your software side you have to avoid reading and writing the file when a different service is about to read/write. Avoiding concurrency isn't a big problem if you have infrequent writes from 3-5 services, but it will be difficult to handle if you have dozens of services that are writing often (which is why you should have a separate file for each user). Also, make sure you clean up old Access Tokens when you are done with them.

  • Thank you for your response, I am actually not persisting the acess_token at all. its fetched used and discarded from memory. And I will keep in mind to specifically delete it from memory. The question is with regards to refresh_token but everything you have said still holds. Apart from writes as it will rarely be overwritten making concurrency simpler
    – shoaib30
    Aug 11, 2021 at 6:21

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.