In this paper, the optimal cyber-attack on remote estimation center in multi nonlinear systems is obtained to study defense strategies in future better. It is assumed that there are F independent nonlinear systems, and each of them has a remote sensor for monitoring. An attacker in the network is considered to exacerbate the specific number of communication channels by generating noise. Because of the capacity limitations in real-world systems, a considered attacker can aggravate at most G of the F communication channels based on its responsibility. The problem is derived as a Markov decision process (MDP) one. Besides, the proof of the existence of an optimal cyber-attack, it is shown that based on the obtained threshold structure of the optimal policy, this problem can be solved for homogeneous models asymptotically, which eases the complexity of computations in real-world applications. In order to check the optimal policy’s performance, numerical results are illustrated.