Like any other cryptanalytic attack, the success rate of a linear attack is expected to improve as more data becomes available. Bogdanov and Tischhauser (FSE 2013) made the rather surprising claim that the success rate of a linear attack may go down with increasing plaintext amount, after an optimal point. They supported this claim with experimental evidence by an attack on SmallPresent-20. Different explanations have been given to explain this surprising phenomenon. In this note, we give

more »

... tative values regarding when this phenomenon can be observed. We conclude that it should not be an issue for attacks in practice except for those with a tiny bias.