The Statistical Disclosure Attack (SDA) is an effective technique for de-anonymizing users in anonymous communication networks. It was presented as a signal detection problem, aiming to distinguish the signal (messaging partners of the targeted user) from the noisy channel (all users). Although the success of signal detection relies on a better understanding of noise properties, the SDA lacks a specific method to analyze the background noise present in the targeted user’s signal, i.e., a noise estimation method. Several noise estimation methods have been proposed in previous studies by utilizing communication rounds in which the targeted user does not participate, called noise rounds. However, these approaches treat these rounds equally without considering their relation with background noise, limiting their effectiveness. In this paper, we propose a novel noise estimation method that weights the noise rounds based on their relation with the background noise present in the targeted user’s signal. We formulate this approach as an optimization problem, where the objective function measures the difference between the current noise estimation and the actual background noise. We minimize this difference by optimizing weights using the Artificial Bee Colony (ABC) algorithm. Our findings indicate that converging noise estimation to the background noise improves the accuracy of the attack.