This paper focuses on the output feedback tracking control of a high-order nonlinear system with denial-of-service (DoS) attacks and exogenous perturbations. A novel neural network (NN)based state estimator is employed to observe the system states during DoS attacks. New nonlinear filters with adaptive regulation are developed to reduce the conservatism of estimator designs and eliminate the impact of unexpected factors such as perturbations and NN approximation errors. Constituting with filtering operation and backstepping control technique, an NN-based secure controller is constructed to suppress the influences of nonlinearities and DoS attacks. The semiglobally uniformly ultimately bounded output tracking results are established by Lyapunov functions on the basis of the estimation and control signals in the cases of DoS attacks, nonlinearities, and perturbations. Comparative results are provided to validate the efficiency of the developed NN-based observation and secure control strategies of a nonlinear system.