Just a regular Joe.

  • 0 Posts
  • 8 Comments
Joined 2 years ago
cake
Cake day: July 7th, 2023

help-circle
  • Encryption will typically be CPU bound, while many servers will be I/O bound (eg. File hosting, rather than computing stuff). So it will probably be fine.

    Encryption can help with the case that someone gets physical access to the machine or hard disk. If they can login to the running system (or dump RAM, which is possible with VMs & containers), it won’t bring much value.

    You will of course need to login and mount the encrypted volume after a restart.

    At my work, we want to make sure that secrets are adequately protected at rest, and we follow good hygiene practices like regularly rotating credentials, time limited certificates, etc. We tend to trust AWS KMS to encrypt our data, except for a few special use cases.

    Do you have a particular risk that you are worried about?


  • Joe@discuss.tchncs.detoSelfhosted@lemmy.worldSecrets Management
    link
    fedilink
    English
    arrow-up
    3
    arrow-down
    1
    ·
    11 hours ago

    Normally you wouldn’t need a secrets store on the same server as you need the secrets, as they are often stored unencrypted by the service/app that needs it. An encrypted disk might be better in that case.

    That said, Vault has some useful features like issuing temporary credentials (eg. for access to AWS, DBs, servers) or certificate management. If you have these use-cases, it could be useful, even on the same server.

    At my work, we tend to store deployment-time secrets either in protected Gitlab variables or in Vault. Sometimes we use AWS KMS to encrypt values in config files, which we checkin to git repositories.


  • wg-quick takes a different approach, using an ip rule to send all traffic (except its own) to a different routing table with only the wireguard interface. I topped it up with iptables rules to block everything except DNS and the wireguard udp port on the main interface. I also disabled ipv6 on the main interface, to avoid any non-RFC1918 addresses appearing in the (in my case) container at all.

    edit: you can also do ip rule matching based on uid, such that you could force all non-root users to use your custom route table.