The Generic Access Network (GAN) developed by 3GPP extends services to unlicensed spectrum. A mobile station (MS) with multiple air interfaces can access 3GPP services through GAN and roam seamlessly between different radio access networks. However, mobility management and resource management are critical issues in GAN. To allocate GAN resources efficiently, we propose Adaptive Keep-alive Interval (AKI). By using AKI, an MS can send less signaling messages. In addition, AKI can increase system utilization by releasing unused resources. In this paper, we develop a mathematical model to evaluate the handoff failure probability of the proposed AKI. The analysis is validated by extensive simulations. The results show that the proposed AKI can reduce handoff failure probability and increase GAN utilization significantly.