safe enough 8-character short unique random string

Question

I am trying to compute 8-character short unique random filenames for, let's say, thousands of files without probable name collision. Is this method safe enough?

base64.urlsafe_b64encode(hashlib.md5(os.urandom(128)).digest())[:8] 

Edit

To be clearer, I am trying to achieve simplest possible obfuscation of filenames being uploaded to a storage.

I figured out that 8-character string, random enough, would be very efficient and simple way to store tens of thousands of files without probable collision, when implemented right. I don't need guaranteed uniqueness, only high-enough improbability of name collision (talking about only thousands of names).

Files are being stored in concurrent environment, so incrementing shared counter is achievable, but complicated. Storing counter in database would be inefficient.

I am also facing the fact that random() under some circumstances returns same pseudorandom sequences in different processes.

Answer 1

Your current method should be safe enough, but you could also take a look into the uuid module. e.g.

import uuid  print str(uuid.uuid4())[:8] 

Output:

ef21b9ad