A joint optimization scheme of both time allocation and mode selection is proposed for defending smart jammer in the Internet of Things (IoT). Specifically, we employ multiple modes to deal with the potential eavesdropping and jamming attacks in wireless sensor networks (WSNs), respectively, as wireless-powered relaying (WPR), ambient backscatter relaying (ABR), friendly jamming (FJ) modes. Assuming a specific frame structure, time allocation between energy harvesting and transmission duration is first optimized for three modes based on transmission and security data maximization. In addition, considering that a single mode may fail to guarantee the performance demand due to the uncertainty of network conditions and adversary attacks, mode selection is carried out for adapting characteristics of different modes. Moreover, according to the available knowledge of jamming detection, for enhancing the robustness in secure transmission, the process of mode selection is further divided into two phases. Specifically, game theory is also utilized to achieve the Nash equilibrium for mode selection. Finally, numerical results verify the effects and exhibit the advantages of the proposed scheme.