In this paper, we study the multistatic cooperative sensing assisted secure transmission via intelligent reflecting surface (IRS). To be specific, a multistatic cooperative sensing scheme is proposed to obtain the angle of arrival and the localization of the eavesdropping target to achieve the beam alignment accurately. Our goal is to maximize the sum secrecy rate by jointly optimizing the association variables of base station (BS) and users, the BS beamforming and the IRS phase shifts, subject to the requirement of target sensing. Due to the coupling of variables and nonconvex objective function, we decompose it into three subproblems and develop an alternative algorithm to solve them iteratively. The association variables are first obtained by successive convex approximation. Then, the BS transmit beamforming can be derived via the semidefinite relaxation. Finally, an alternating direction method of multiplier is adopted for the IRS phase-shift design. Simulation results indicate that the multistatic cooperative sensing via IRS can enhance the sensing performance and guarantee the security.