Abstract: Uncrewed aerial vehicle (UAV)-assisted integrated sensing and communication (ISAC) is a key enabler of next-generation wireless networks. However, the open nature of low-altitude propagation channels exposes these systems to severe physical layer security vulnerabilities. Existing research often oversimplifies these security environments, failing to fully address complex scenarios in which spatially distributed and functionally heterogeneous malicious entities coexist. Motivated by this gap, we propose a novel robust active reconfigurable intelligent surface (RIS)-assisted ISAC system architecture. Specifically, addressing the lack of a direct line-of-sight link between the base station and sensing targets, an active RIS is required to establish the necessary sensing and communication links. The proposed framework comprehensively accounts for three distinct classes of threats: ground passive eavesdroppers intercepting confidential information, aerial active jammers emitting malicious signals to disrupt network functionalities, and aerial hybrid malicious entities capable of executing simultaneous eavesdropping and jamming attacks. To effectively counter these diverse threats, a dual-defense strategy is proposed. First, an artificial noise (AN) jamming mechanism is employed at the base station to deliberately degrade the reception quality of passive eavesdroppers without affecting cooperative users. Second, an active RIS enhancement mechanism is utilized to simultaneously adjust phase shifts and amplify the reflected signal amplitude to boost the desired signal power to combat malicious active jamming. For sophisticated hybrid entities, both AN and active RIS mechanisms are applied synergistically. Based on this system model, we formulate a joint optimization problem for the transmit beamforming matrix at the base station and the reflection coefficient matrix at the active RIS. The objective is to maximize the sensing signal-to-clutter-plus-noise ratio (SCNR) subject to strict inequality constraints on the communication and security signal-to-interference-plus-noise ratios (SINRs). Owing to the highly coupled nature of the optimization variables and the presence of double-continuous variables (amplitude and phase) at the active RIS, the formulated problem is highly nonconvex and mathematically intractable. To solve this problem without incurring the exponential computational complexity of exhaustive search methods, we propose an algorithm framework based on block coordinate descent. The original high-dimensional joint optimization problem is decoupled into two tractable subproblems: transmit beamforming optimization and active RIS reflection coefficient matrix optimization. These subproblems are iteratively solved using convex optimization techniques, including the majorization-minimization algorithm and semidefinite relaxation method. Theoretical analysis demonstrates that the proposed algorithm strictly controls the computational overhead within the polynomial time. The simulation results verify the effectiveness and robustness of the proposed joint optimization algorithm. The results demonstrate the superiority of the active RIS architecture over conventional passive RIS baselines. Specifically, under stringent security SINR constraints, ablation studies reveal that the proposed AN jamming and RIS enhancement mechanisms independently improve the sensing SCNR by approximately 49.75% and 22.58%, respectively. Furthermore, the complex trade-offs between communication and sensing functionalities and the robustness of the system against channel uncertainties are thoroughly analyzed. Finally, while the current framework assumes the known identities of malicious entities, future work will focus on relaxing this assumption to investigate the joint design of blind recognition for unknown targets based on ISAC echo characteristics.