Is it safe to rename file.gpg to `md5sum file`?

[Ben Staude]

Am 05.12.2012 18:59, schrieb Max Parmer:
> On Tue, Dec 4, 2012 at 12:03 PM, sben1783 <sben1783 at yahoo.de> wrote:
>> Yes, I meant to use the MD5 checksum of the original file, not its
>> original name. I'm still interested whether this would be "insecure"?

> If by insecure you mean, "could lead to exposing the contents of the file"
> or "could reveal my passphrase" that would depend (in part) on the size and
> contents of the file (i.e. very short files would less time consuming to
> brute force, files with very regular formats would be quicker to brute
> force, etc.) and the symmetric cipher used.

Yes, that's exactly what I meant with "insecure". So I learn from your 
statement to better not use the md5 hashes.

> Revealing the plaintext of some files could be fairly significant with the
> default symmetric cipher for GPG is CAST-128 which is potentially subject
> to key recovery via a chosen plaintext attack. AES doesn't have any
> presently known vulnerability of that sort.

While browsing for recommandations on what algorithm to use, I didn't 
come across the "vulerability" you mention above. I don't really 
understand what you're saying, but anyway plan to use AES because it a) 
seems to be the algorithm "more state of the art" and b) uses MDC by 
default.

> If you just need a unique key to refer to the file, you're already storing
> the source path in the "summary" file your tool generates. If you just need
> a guaranteed unique identifier for each file (because, say, you're storing
> them all flatly in a single directory), I would just hash the path (which
> is apparently not sensitive data, as you seem to be storing it in plaintext
> in the summary file) as it's guaranteed to be unique per-system.

Well I do *not* want to reveal my private paths/filenames in the remote 
backup location. I won't upload the summary file as plaintext, but maybe 
encrypted as contents.gpg or the like. So I need another identifier for 
each file and some sort of mapping. That's why I came up with the md5sum 
of the files contents in the first place - I already have the mapping 
table (the summary file). If that's no good idea, I will probably just 
use a GUID for each file and create a separate mapping table (which also 
won't get uploaded without encryption:)

If I wanted to have a fallback for loosing the mapping table, would 
there be a sane way to encrypt the filename with gpg? That way I could 
decrypt it in case I loose the mapping table (which isn't possible with 
the GUID solution). I tried

echo '/path/to/original/filename' | gpg --armor -c

but the result contains newlines and slashes which isn't good for use as 
filenames. There's no option like "--armor-only-alphanumeric"...?

Thank you very much
Ben