Thermal WIMPs are well-motivated dark matter candidates, and are being searched by various experiments. However any clear signature is not observed yet and the resultant constraints on thermal WIMP models get stronger and stronger. In particular, the constraints of recent direct detection experiments exclude a large region of parameter space below TeV scale. In this talk, we propose a model of pseudo-Nambu-Goldstone dark matter, and we show that the dark matter can naturally avoid the strong constraints of dark matter direct detection. In addition to detailed calculation of elastic cross section of dark matter, the other ways of detecting the dark matter are also discussed.