duplicity-talk
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Duplicity-talk] Manifest stores SHA1 hash of files, checked before


From: Kenneth Loafman
Subject: Re: [Duplicity-talk] Manifest stores SHA1 hash of files, checked before restore?
Date: Thu, 14 Jul 2011 07:44:01 -0500

On Thu, Jul 14, 2011 at 6:43 AM, <address@hidden> wrote:
On 14.07.2011 12:19, Chris Poole wrote:
> On Thu, Jul 14, 2011 at 9:38 AM,  <address@hidden> wrote:
>> On 13.07.2011 17:53, Chris Poole wrote:
>>> (Thus, it's very important to sign
>>> backups being stored in untrusted locations.)
>>
>> It is provided the public key used is published somewhere or in other ways available to a possible attacker. If you create a keypair just for your backup and keep it on the backup machine and in your secure storage (for restoring) you don't necessarily need it.
>>
>> On the other hand. Currently duplicity needs a private key to work reliably, so signing to it does no harm and can be seen as an extra lock for an intruder to pick. see http://bugs.launchpad.net/duplicity/+bug/687295
>
> Thanks. I'm going to get used to signing my backups. I don't use cron
> to do them for me anyway.
>
> What I find annoying is that Duplicity asks me for my passphrase (when
> doing an incremental backup) 3 times. Surely once is enough, to
> decrypt my private key? (Using the same Key ID to encrypt and sign my
> backup.)
>

latest duplicity has the possibility to define env var SIGN_PASSPHRASE and PASSPHRASE. this way you don't have to input them manually.

there is no code to compare signing vs. encryption key, so they are asked for separately. I am not sure if the double input to ensure correctness is a wise decision. i would plead to have it putted in and if it is wrong gpg will complain later on.

@ken: is the doublecheck routine really necessary?

eventually. i just had a look at the corresponding code duplicity-bin::get_passphrase. with the latest duplicity you should be asked two times ("Input/Retype") for each key (Signing/Encryption). Isn't that so? You could post an obfuscated output log of a run with '-v9' to show what happens.

Verifying the password is not absolutely necessary, but saves some time if the user inputs the wrong password.

It should never double verify.  I'd like to see the first 200 lines of the log from a -v9 run to see what triggered it.

...Ken


reply via email to

[Prev in Thread] Current Thread [Next in Thread]