Meet-in-the-middle attacks on 10-round AES-256

Meet-in-the-middle attack on AES is proposed by Demirci and Selçuk at FSE 2008, and improved greatly by Dunkelman et al. at ASIACRYPT 2010 and Derbez et al. at EUROCRYPT 2013 with various time/memory/data tradeoff techniques. At FSE 2014, Li et al. give the most efficient attack on 9-round AES-256 based on a 5-round meet-in-the-middle distinguisher. In this paper, we revisit Demirci and Selçuk’s attack and present the first 6-round meet-in-the-middle distinguisher on AES-256 using the differential enumerate and key-dependent sieve techniques. Based on this distinguisher, we propose the first attack on 10-round AES-256 in the single-key model except biclique attack. Moreover, we can further reduce the data complexity by using several distinguishers in parallel and reduce the memory complexity by dividing the whole attack into a series of weak-key attacks. Finally, we can achieve the attack with a data complexity of $$2^{111}$$2111 chosen plaintexts, a time complexity of $$2^{253}$$2253 10-round AES encryptions and a memory complexity of $$2^{211.2}$$2211.2 AES blocks.