The information-centric network (ICN) provides network protocols oriented to information itself, including a content-centric subscription mechanism and semantic-led naming, routing, and caching strategies. It has shown great potential in solving attacks on current IP address based network. This paper aims to propose a smart firewall model for ICN, and to build a firewall based on a semantic inference algorithm to isolate content. The ICN firewall module uses the fog computing paradigm to sense content threats from ICN, and generates customized filtering strategies for different contents. On the basis of analyzing the types of ICN attacks and the development of the fog computing architecture in ICN, this article introduces the fog-based ICN firewall model from three aspects: the overall structure of content defense, the host-oriented defense fog model, and the network-oriented defense fog model. This article also proposes an ICN-oriented detection and defense mechanism in order to alleviate the interest flooding attacks. Finally, by building the ndnSIM network simulation platform, this article evaluates the ICN cache hit rate and network communication delay, and verifies the feasibility and efficiency of the proposed fog computing-based ICN firewall module and defense algorithm.