In this study, we investigate the time-varying finite-time synchronization of attack-induced uncertain neural networks. We propose a novel time-varying finite-time control scheme to achieve network synchronization. The proposed control employs a time-varying control gain that remains positive and bound when averaged over a specific time period. Moreover, our method takes into consideration various factors that can induce uncertainties in neural networks, including both unconscious internal and external influences, as well as deliberate attacks. By leveraging the classical Lyapunov method, we derive a new finite-time lemma and establish sufficient criteria for ensuring that all neural states of the networks synchronize to a desired state within the settling time. Through numerical simulations, we validate the efficacy of our proposed method.