<p>This article proposes an improved command-filter-based predefined-time consensus tracking control strategy for nonlinear multi-agent systems (MASs) subject to false data injection (FDI) attacks. Such attacks targeting sensors and actuators compromise the inputs, outputs, and states of MASs. To address this challenge, an observer is designed to estimate the compromised system states under attacks. Furthermore, fuzzy logic systems (FLSs) are leveraged to approximate unknown nonlinearities and deceptive injected signals. To overcome the unknown control gain problem arising from actuators attacks, a Nussbaum-type function is incorporated into the controller design. Then, within the backstepping framework, an adaptive command filter is designed by combining the dynamic adaptive technique. This design effectively mitigates the “complexity explosion” issue inherent in traditional backstepping while bolstering the robustness of the MASs. Concurrently, an improved predefined-time compensation mechanism is devised to counteract the effects of filtering errors. By synthesizing adaptive backstepping technique with predefined-time control, an adaptive consensus controller is developed. The proposed controller can realize the convergence of consensus error, even in the presence of dual attacks on actuators and sensors. Ultimately, the feasibility of the suggested methodology is verified via both numerical and practical simulation.</p>