Affected versions of csrf-lite
are vulnerable to timing attacks as a result of testing CSRF tokens via a fail-early comparison instead of a constant-time comparison.
Timing attacks remove the exponential increase in entropy gained from increased secret length, by providing per-character feedback on the correctness of a guess via miniscule timing differences.
Under favorable network conditions, an attacker can exploit this to guess the secret in no more than (16*18)288 guesses, instead of the 16^18 guesses required were the timing attack not present.
Update to version 0.1.2 or later.