Information hiding technology is an important part of network security, and it also faces the challenge brought by new technology. The traditional method of hiding network sensitive links can be easily exposed by attackers using link prediction algorithms. In order to strengthen network information security, a network information hiding method against link prediction is proposed. This method can disturb the network to a certain extent, so that the attacker cannot discover hidden sensitive links through link prediction, so as to achieve the purpose of protecting privacy. To select the optimal perturbed links, a set of genetic algorithm is designed. Both antagonistic effect and perturbed degree are considered in the fitness function. The final perturbed links obtained through multiple iterations can not only reduce the score of sensitive links in the link prediction algorithm, but also maintain the stability of the network structural features. The effectiveness of the proposed algorithm is verified by comparing with the baseline method on real data sets.