Passive testing has proved to be a powerful technique for protocol system fault detection by observing its input/output behaviors yet without interrupting its normal operations. To improve the fault detection capabilities we propose a backward checking method that analyses in a backward fashion the input/output trace from passive testing and its past. It effectively checks both the control and data portion of a protocol system, compliments the forward checking approaches, and detects more errors. We present our algorithm, study its termination and complexity, and report experiment results on a real protocol SCP.