Due to the challenging nature of nonlinearity and time-varying behaviors in TCP dynamics, traditional active queue management (AQM) algorithms suffer from poor robustness under dynamic network scenarios. To address the above problem, a novel congestion control scheme named FSM-AQM is proposed via the fuzzy sliding mode control, which is robust against uncertainties and time-varying parameters of the system. The guidelines to design FSM-AQM are presented, including sliding surface selecting, sliding controller design and stability analysis. As for the reachable issue, an enhanced reaching law based on fuzzy logic is designed to decrease the high-frequency chattering and improve the dynamic performance. Simulation comparisons with PI, PID and REM in NS2 demonstrate that our proposed algorithm achieves fast convergence of queue length to the target with small jitter. The results also indicate that FSM-AQM is robust against disturbances and outperforms traditional AQM schemes significantly.