<p>This paper investigates the mean-square bounded consensus of a kind of nonlinear multi-agent systems (MASs) with time-varying. A novel dual-channel stochastic deceptive attack scheme is designed for this system (MASs) by utilizing directional network structures. In this new scheme, both sensor-controller (S-C) and controller-actuator (C-A) channels with distinct tampered signals are considered during variable impulsive control intervals. To counter these attacks, an adaptive secure impulsive coordination strategy has been developed for the nonlinear MASs with time-varying and dynamic switching capabilities. Based on Lyapunov’s stability theorem, methods of matrix analysis, and linear matrix inequality techniques, some adequate conditions have been established to guarantee consensus in the mean-square sense with bounded errors for the system if the stochastic selection of compromised channels obeys Bernoulli distributions. Two simulation cases are given to address the effectiveness of the derived results.</p>