3 ms·
> Compared to Google Authenticator app, YubiKey (a) makes hardware-based OTPs as opposed to time-based OTPs (does that offer stronger security?) If I understan
by e12e 11y ago
> Compared to Google Authenticator app, YubiKey (a) makes hardware-based OTPs as opposed to time-based OTPs (does that offer stronger security?)
If I understand the various OTP implementations correctly, they are essentially the same, trading a time-counter for use-counter.
The non-time OTP will give you ds1, ds2, ds3 -- where dsX ("derived secret X") is derived from shared-secret+X, while the time-based OTP will give you dsT1, dsT2 ... derived from shared-secret+TimestampX.
The former is a little more secure, as you need to allow for some drift. And having a proper 128 bit shared secret is probably more secure than something derived from a crackable password/pass phrase.
But AFAIK they're both vulnerable to the shared secret being compromised on the server-side and/or cracked (if it's possible to brute force).
I don't recall the exact details of the TOTP-spec, but in general I think the derived-secret/otp essentially boils down to: truncate(hmac(shared-secret,counter)) where counter is either a timestamp, or a sequence.
If the counter is actually a sequence of used passwords(1,2,3...N), the server can store just N passwords (usable for N logins), and cross them off after use -- so no need to store the shared secret on the server. For TOTP the server needs the shared-secret in order to generate a few +/- values to match against what the client supplied.
In both cases you could use a sniffed one-time password to try and guess the shared-secret, if you can know/guess the counter. The risk for a 128 bit random key is obviously somewhat less than if the shared secret is "hunter2" (modulus stretching etc).