4 ms·

Cracking Passwords in the Cloud: Breaking PGP on EC2 with EDPR

42 points by edw519 17y ago

tptacek 17y ago
This isn't "breaking PGP", it's guessing PGP passphrases. If they had guesses a PGP key using a cluster of virtual machines, that would be news.
- cperciva 17y ago
  it's guessing PGP passphrases Also, if you can brute-force a PGP passphrase using a cluster of EC2 instances, it's trivial to brute-force it in hardware (using FPGAs or, ideally, ASICs). If you want to protect something with a passphrase, use scrypt!
  - tptacek 17y ago
    No matter how you hash them, passphrases are still pretty secure. If you can type your PGP passphrase very quickly, you're doing it wrong.
    - cperciva 17y ago
      passphrases are still pretty secure ... at least until someone comes along with a microphone (or a cell phone) and exploits the acoustic side channel. :-)
- ax0n 17y ago
  Hear, hear! They didn't crack or break PGP at all. They still needed the password-protected secret key to pull it off. It's not much different than attempting to brute-force anything else, including login:password pairs.
bcl 17y ago
The passphrase is the weak point in PGP. I suppose some people treat it like a password, which is the wrong approach. It should be much longer and harder than a normal password. Their analysis does demonstrate how cost-effective it has become to crack low-complexity passwords.
sharms 17y ago
They would also need a way to validate the data unless the encryption was symmetric, I don't think they have tested this out on data they didn't already have.
- Kadin 17y ago
  They were brute-forcing a PGP-encrypted ZIP archive, which would have been symmetric encryption. These kinds of archives were produced by the PGP Desktop application (at least in earlier versions; I don't know if it still does) and were functionally similar to using "gpg -c", although the actual file format is/was different and incompatible. Had the guy used a public/private keypair for encryption rather than a password, I suspect the work needed to bruteforce it would be much greater. But that doesn't look like the situation.
papyromancer 17y ago
The cost analysis is great ;) Just by following the guidelines of using a complex character set and decent length for your passphrases is sufficient to make the cost of cracking it brute force ridiculously expensive. Just watch out for the wrench exploit ;P

42 points by edw519 17y ago

This isn't "breaking PGP", it's guessing PGP passphrases. If they had guesses a PGP key using a cluster of virtual machines, that would be news.

cperciva 17y ago
it's guessing PGP passphrases Also, if you can brute-force a PGP passphrase using a cluster of EC2 instances, it's trivial to brute-force it in hardware (using FPGAs or, ideally, ASICs). If you want to protect something with a passphrase, use scrypt!
- tptacek 17y ago
  No matter how you hash them, passphrases are still pretty secure. If you can type your PGP passphrase very quickly, you're doing it wrong.
  - cperciva 17y ago
    passphrases are still pretty secure ... at least until someone comes along with a microphone (or a cell phone) and exploits the acoustic side channel. :-)
ax0n 17y ago
Hear, hear! They didn't crack or break PGP at all. They still needed the password-protected secret key to pull it off. It's not much different than attempting to brute-force anything else, including login:password pairs.

bcl 17y ago

The passphrase is the weak point in PGP. I suppose some people treat it like a password, which is the wrong approach. It should be much longer and harder than a normal password. Their analysis does demonstrate how cost-effective it has become to crack low-complexity passwords.

sharms 17y ago

They would also need a way to validate the data unless the encryption was symmetric, I don't think they have tested this out on data they didn't already have.

Kadin 17y ago
They were brute-forcing a PGP-encrypted ZIP archive, which would have been symmetric encryption. These kinds of archives were produced by the PGP Desktop application (at least in earlier versions; I don't know if it still does) and were functionally similar to using "gpg -c", although the actual file format is/was different and incompatible. Had the guy used a public/private keypair for encryption rather than a password, I suspect the work needed to bruteforce it would be much greater. But that doesn't look like the situation.

papyromancer 17y ago

The cost analysis is great ;) Just by following the guidelines of using a complex character set and decent length for your passphrases is sufficient to make the cost of cracking it brute force ridiculously expensive. Just watch out for the wrench exploit ;P