Abstract: In conventional integrated sensing and communication （ISAC） systems， spectrum sharing between communication and sensing functions， combined with the broadcast nature of signals， renders ISAC systems vulnerable to malicious attacks by eavesdroppers. To address the security challenges in information transmission， we investigated a secure transmission scheme for an intelligent reflecting surface （IRS）-assisted ISAC system based on rate-splitting multiple access （RSMA）. First， a multi-user downlink communication transmission model is established for the IRS-assisted ISAC system based on RSMA. When an eavesdropper attempts to intercept confidential information intended for legitimate users， the rate-splitting feature of RSMA is leveraged to jointly design beamforming for common and private information. This design induces additional interference from the common information， thereby degrading the ability of the eavesdropper to decode private information. Meanwhile， the reflective beamforming gain provided by the IRS is exploited to enhance the security of ISAC communications. Next， an optimization problem is formulated with the objective of maximizing the secrecy rate while satisfying constraints on the maximum transmission power， IRS reflection phase shifts， and beam gain in the sensing target direction. The problem is structured to jointly optimize the beamforming vectors for common and private information， the rate-splitting vector， and the IRS phase shift matrix. An alternative optimization approach is then employed to address this non-convex problem. By eliminating the non-convex rank-one constraint using the semi-definite relaxation （SDR） algorithm， the successive convex approximation （SCA） method is applied to approximate the two non-convex subproblems into convex optimization problems. Simulation results show that the proposed scheme effectively enhances the secrecy rate while meeting sensing performance requirements and confirm the advantage of optimizing IRS reflection phase shifts in ISAC.