Game Theoretic Path Selection to Support Security in Device-to-Device Communications

摘要

1 Device-to-Device (D2D) communication is expected to be a key feature supportedudby 5G networks, especially due to the proliferation of Mobile EdgeudComputing (MEC), which has a prominent role in reducing network stressudby shifting computational tasks from the Internet to the mobile edge. Apartudfrom being part of MEC, D2D can extend cellular coverage allowing users toudcommunicate directly when telecommunication infrastructure is highly congestedudor absent. This significant departure from the typical cellular paradigmudimposes the need for decentralised network routing protocols. Moreover, enhancedudcapabilities of mobile devices and D2D networking will likely result inudproliferation of new malware types and epidemics. Although the literature isudrich in terms of D2D routing protocols that enhance quality-of-service and energyudconsumption, they provide only basic security support, e.g., in the form of encryption. Routing decisions can, however, contribute to collaborative detectionudof mobile malware by leveraging different kinds of anti-malware softwareudinstalled on mobile devices. Benefiting from the cooperative nature ofudD2D communications, devices can rely on each other's contributions to detectudmalware. The impact of our work is geared towards having more malware-freeudD2D networks. To achieve this, we designed and implemented a noveludrouting protocol for D2D communications that optimises routing decisionsudfor explicitly improving malware detection. The protocol identifies optimaludnetwork paths, in terms of malware mitigation and energy spent for malwareuddetection, based on a game theoretic model. Diverse capabilities of networkuddevices running different types of anti-malware software and their potentialudfor inspecting messages relayed towards an intended destination device areudleveraged using game theoretic tools. An optimality analysis of both Nashudand Stackelberg security games is undertaken, including both zero and nonzeroudsum variants, and the Defender's equilibrium strategies. By undertakingudnetwork simulations, theoretical results obtained are illustrated through randomlyudgenerated network scenarios showing how our protocol outperformsudconventional routing protocols, in terms of expected payoff, which consistsudof: security damage inflicted by malware and malware detection cost.