nachonavarro / balloon-hashing Goto Github PK

On page 7 of the paper, you can see that cnt starts at 0, it is still 0 when it is passed to the first hash function in the expand stage. It is then incremented according to s_cost. When cnt is used during the mix stage, its value is greater than its initial value.

In this implementation, cnt is initialised to 1, a reference is then passed to the expand function. In Python, integers are immutable so when it is incremented a copy is made and that copy is incremented, instead of incrementing the original. When cnt is passed to the first hash function in the expand stage, its value is 1. It is then incremented according to s_cost, but the accumulated value is thrown away at the end of the expand function, and when cnt is passed to mix, it is still its initial value.

As cnt is used in the security proof this divergence in the implementation should be considered a severe error. I suggest you stop breaking the balloon function up into sub functions, as how Python passes primitives to functions is confusing and is leading to mistakes.