Improve Neural Distinguishers of SIMON and SPECK

Table 3

Comparison with Algorithm 1 and other methods.

Block cipher	Source of neural distinguisher (input difference)	Rounds	Input difference	Accuracy (%)
SIMON32/64	Reference [17]^1	9		59.07
	Algorithm 1	9		59.77
SIMON48/96	Reference [17]^1	9		50.22
	Reference [10]^2	10		53.49
	Algorithm 1	10		57.89
SIMON64/128	Reference [17]^1	10		58.61
	Algorithm 1	11		59.72

¹In [17], Abed et al. constructed differential characteristics of SIMON. We train neural distinguishers by choosing the input differences in [17]. ²We train neural distinguishers using Benamira et al.'s method presented in [10].