<p>This paper examines the bipartite bounded consensus of multiagent systems (MASs) connected by signed graphs. The considered MAS includes a virtual leader and multiple followers with nonlinear dynamics, where communication link weights between neighboring agents can be negative. To achieve consensus, impulsive control depending on neighbor information is utilized. However, this control may be subjected to deception attacks. To optimize control efficiency by reducing frequency and shortening consensus time, a self-triggered mechanism that determines impulsive instants with variable intervals is proposed. Utilizing graph theory, linear matrix inequality (LMI), and the Lyapunov functional method, conditions for achieving bipartite bounded consensus and the consensus error bound are provided. This study reveals that the graph topology, attack probability, and the maximum value of impulsive intervals are key factors affecting the consensus. Numerical simulations validate the theoretical findings. A comparison of strategies with fixed and self-triggered impulsive intervals highlights the effectiveness of the self-triggered scheme.</p>