Discarding the Endpoints makes the Cryptanalytic Time-Memory Trade-Offs even Faster

Cryptanalytic time-memory trade-offs were introduced by Hellman in 1980 in order to perform key-recovery attacks on cryptosystems. A major advance was presented at Crypto 2003 by Oechslin, with the rainbow table variant that outperforms Hellman’s seminal work. This paper introduces the fingerprint tables, which drastically reduce the number of false alarms during the attack compared to the rainbow tables. The key point of our technique consists in storing in the tables the fingerprints of the chains instead of their endpoints. The fingerprint tables provide a time-memory trade-off that is about two times faster than the rainbow tables on usual problem sizes. We experimentally illustrate the performance of our technique, and demonstrate that it is faster than Ophcrack, a Windows LM Hash password cracker considered so far to be the fastest one ever implemented.