what s the longterm security margin of of a winzip or Winrar file encrypted with aes-256 and a password consisting of 64 completely random characters set a-z 0-9 and other printable special characters)?

How long will such a file be safe against an enemy with a large GPU cluster?

My own calculation is that the possible number of passwords is 96^64.

I have searched but all the amazing success stories turning up are about bruteforcing of weaker MD5 and Windows passwords.

Very little specific about cracking of Winzip, Winrar or PGP protected files on GPUs where the password is longer than 10 characters.

Does it mean that Winzip, Winrar and PGP encrypted files are not easily subject to the same attacks?

Or does it mean that we'll have to wait a few years before this security margin is broken by evolution in computational power?

  • Password based encryption is only as good as the password that is used. If you use a 5 character password with Winzip, Winrar, or PGP... you're gonna have a bad time like anything else. If it's truly random printable characters once you get to the 12-14 character range you're looking at a large password space that GPUs still haven't quite reached when it comes to cracking. If you're not seeing successful cracks in higher password lengths then it might mean that their password scheme isn't flawed like it could be with other software. – RoraΖ Aug 19 '14 at 11:52

Modern versions of WinZIP and WinRAR employ fairly slow key derivation in order to combat password guessing attacks. In particular, WinRAR uses scheme where number of effective SHA-1 iterations depends on length of the password (longer passwords yield more iterations), and WinZIP uses PBKDF2-based function, IIRC.

"Fairly slow" is not a precise metric. For WinZIP actual speed will depend on ZIP file format; older formats used weak encryption and weak key derivation; later formats are much better in that respect. If you need some numbers, you can assume one modern GPU can deliver about one million passwords per second. For WinRAR that figure is much lower and will likely be below 50'000 passwords/sec per GPU.

Using those figures you can actually estimate what length of a random password is sufficient for your purposes (e.g. how long it should resist attacks).

I'd also like to note that 64-char random password is an overkill. 20 characters will already give you more than 128 bits of strength which is more than sufficient (encryption in WinRAR is AES-128 anyway). If you want 256 bits of strength then 40 characters is sufficient.

| improve this answer | |

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.