Dynamic average consensus (DAC) algorithms are widely applied in distributed multi-agent systems. However, most existing DAC algorithms require explicit state transmission, which inevitably exposes private reference signals to external eavesdroppers. This paper proposes a novel privacy-preserving event-triggered DAC algorithm based on agent decomposition. Each agent is virtually split into an observable \(\alpha \) -subsystem for event-triggered communication and a concealed \(\beta \) -subsystem evolving only via local dynamics, thereby breaking any direct mapping between transmitted messages and sensitive reference signals. We prove that all substates converge exponentially to a bounded neighborhood of the time-varying average, derive an explicit ultimate tracking error bound, and establish a strictly positive lower bound for inter-event intervals to rigorously exclude Zeno behavior. Under an information-theoretic indistinguishability criterion, the proposed algorithm guarantees strict privacy of each agent’s reference signal against an external eavesdropper with full knowledge of the network topology, algorithm parameters, and all intercepted messages. Unlike many privacy-preserving DAC methods that remain vulnerable to eavesdropping, impose restrictive signal assumptions, or rely on neighbor cooperation for sum-conservation constraints, the proposed algorithm achieves privacy protection without such limitations. Numerical simulations are provided to verify the effectiveness of the proposed algorithm.
