Observations Concerning the probability of the existence of annihilators for balanced boolean functions

LFSR-based stream ciphers with nonlinear filters or combiners are susceptible to algebraic attacks using linearization methods to solve an overdefined system of nonlinear equations. And this process is greatly enhanced if the filtering or combining function has a low degree annihilator. To prevent such an attack, one would choose the parameters of that function so that the degree of its annihilator becomes large enough. As computing power is continuously increasing, a choice that seems secure today, becomes insecure tomorrow. Therefore, a tool is needed to estimate the probability of the existence of annihilators for balanced boolean functions with parameters that are beyond the current computing power. Based on experimental and calculational observations, we give in this paper an almost exact estimate of that probability, which represent a great improvement over the upper bound previously known.