Due to the wide variety of attacks and the difficulties of developing an accurate statistical model of host features, the structure of the watermark detector is derived by considering a simplified channel model. In this paper, we present a fragile watermarking based on block-mapping mechanism which can perfectly recover the host image from its tampered version by generating a reference data. By investigating characteristics of watermark detector, we make an effective analysis such as fragility against robustness measure and distinguish its property. In particular, we derive a watermark detector structure with simplified channel model which focuses on the error probability versus watermark-to-noise-ratio curve and describes a design by calculating the performance of technique, where attacks are either absent or as noise addition.