The problem of tracking the expected market demand and suppressing the bullwhip effect is investigated for an unknown nonlinear supply chain systems (SCS) subject to a denial-of-service (DoS) attack on the commodity output. The supply chain system is first transformed into a dynamic data model using a dynamic linearisation approach. Then, considering that the DoS attack follows a Bernoulli distribution and occurs randomly in the transmission channel between the sensor and the control centre, the lost commodity output due to the DoS attack is replaced by the estimated value of the historical data at the previous moment. The model-free adaptive defensive control algorithm is then designed based on the raw material orders and commodity output of each sub-chain under the DoS attack, and the convergence of the closed-loop system under the defensive control algorithm is demonstrated by theoretical analysis. Finally, the effectiveness of the defensive control algorithm is verified by simulation.