Abstract: In response to the challenges posed by netted radars to traditional interference performance, a drone-based cooperative deception interference method is presented based on the reinforcement learning principles. Initially, leveraging the deceptive interference principles of netted radars, the optimization problem of cooperative control strategy for drone swarm system is formulated as a distributed Markov decision process. Subsequently, a reward function based on artificial potential fields is devised, and a dynamic exploration strategy is introduced to enhance algorithmic efficacies. Furthermore, integrated with a multi-head attention mechanism, the group collaboration capabilities are bolstered, which effectively mitigates the problems such as diminished single-drone precision, sluggish algorithmic convergence, and challenges in multi-drone coordination in cooperative deception interference scenarios. Ultimately, simulation results substantiate that the proposed method outperforms the traditional multi-agent twin delayed deep deterministic policy gradient algorithm in disrupting netted radars, and notably enhances interference performance.