Although various feedback control algorithms are proposed for wall turbulence, most efficient control schemes require the spanwise wall shear stress, which is difficult to measure in physical experiments. In the present study, a new control scheme based on the streamwise wall shear stress is developed with the aid of genetic algorithms. Local blowing/suction at the wall was employed as the control input, and its distribution was optimized in order to minimize the wall skin friction. It is found through direct numerical simulation of turbulent channel flow that about 12% drag reduction is obtained by using a control scheme, which gives asymmetric distribution of the control input in the spanwise direction. The control input becomes out of phase with the vortices because of the drastic change of the streaky structures in the vicinity of the wall.