Integrated sensing and communication (ISAC) is recognized as a promising technology with great potential in saving hardware and spectrum resources, since it simultaneously realizes radar detection and user communication functions in the fully-shared platform. Employing reconfigurable intelligent surface (RIS) in ISAC systems is able to provide a virtual line-of-sight (LoS) path to conquer blockage problem as well as introduce new degrees of freedom (DoFs) to further enhance system performance. Nevertheless, the multiplicative fading effect of passive RIS limits its applications in the absence of direct links, which promotes the development of active RIS. In this paper, we consider an active RIS-assisted ISAC system and aim to jointly design the transmit beamformer, the active RIS reflection and the radar receive filter to maximize the radar output signal-to-noise ratio (SNR) while guaranteeing pre-defined signal-to-interference-plus-noise ratios (SINRs) for communication users. To solve for this non-convex problem, an efficient algorithm is developed by leveraging the techniques of block coordinate descent (BCD), Dinkelbach's transform and majorization-minimization (MM). Simulation results verify the significant advancement of deploying active RIS in ISAC systems, which can achieve up to 32dB radar SNR enhancement compared with the passive RIS-assisted ISAC systems.