We use the multi-epoch z'-band data (Subaru/Suprime-Cam; 2008-2011/6-epoch) to identify low luminosity type-1 AGN at high redshift by their variability in SXDS/UDS field. These low-luminosity AGN candidates are not detected in the existing XMM-Newton X-ray image (~10^{-15} erg/s/cm^2 in the 0.5–2 keV band; Ueda et al. 2008). To broadly constrain the redshift range of the sample, we applied the color-selection, such as BzK, DRG, LBG, to find the high redshift candidates assuming that their colors at longer wavelength are dominated or significantly contributed by stellar light. Consequently we find there are about 600 candidates that have less than 10^8 Msun Black Hole mass and exist in the high redshift (z > 1). We then conduct spectroscopic follow-up observation (Keck/DEIMOS) for a selected small 'pilot' sample to confirm that they are broad-line AGN and obtain their redshift and BH mass. As a result, we find that from 23 observed samples, 5 have MgII (2800A) broad line and one has CIV (1550A) broad line, both of which are signature of AGNs, and find that they have less massive black hole (less than 10^8Msun) at the high redshift (z > 1). These results suggest that AGN variability and host-galaxies' color selection is effective to identify less massive black holes in the high redshift faint galaxies even below the current X-ray sensitivity limit. In this poster, I describe the methods, multi-epoch photometry results, and the results of spectroscopy in detail. And I also show the preliminary results of AGN variability research using Hyper Suprime-Cam Subaru Strategic Program data.