The anisotropic dust torus is invoked in unification theory of Active Galactic Nuclei (AGNs) to account for the observational dichotomy, all AGNs are proposed the same object or they born equal but viewed from different lines of sight. However, little is known about the dust in the circumnuclear torus of AGN. There is evidence suggesting that the size and composition of the dust in AGN may differ substantially from that of the Galactic interstellar dust, as reflected by the flat/``gray'' (or steep, SMC-like) extinction curves. Besides, the anomalous silicate emission or absorption features observed respectively in type 1 and type 2 AGN suggests that the AGN silicate grains were probably not ``born equal''. To address and to clarify the observed diversity among AGNs dichotomy, the dust modeling is required to perform on a large and well defined AGN sample, taking care of various dust composition and dust grain size. By modeling theSpitzer/IRS spectra of 140 AGNs, we obtain constraints over the silicate composition and size. I will introduce the simple approach we take in dust modeling and discuss the implication of our current results.