TOFU: Trust-Based Task Offloading in UAV-as-a-Service

This work proposes a trust evaluation scheme, TOFU, for task offloading in UAV-as-a-Service. Multiple UAV owners participate and collaboratively offer UaaS to the end-users. In the process, a UAV with limited resources offloads the task to other UAVs with adequate resources. However, not all the UAVs are trustworthy to offload the task. In such a case, we need to evaluate the trust of those UAVs to whom another UAV wishes to offload the task. The proposed trust evaluation scheme, TOFU, consists of three components – (i) residual energy, (ii) data transmission rate, and (iii) reputation of a UAV. The residual energy and data transmission are the pre-determined components retrieved directly. On the other hand, we need to compute the reputation of a UAV specific to the UaaS environment. We consider the resources available with a UAV to determine its initial reputation. Further, we apply the debit-credit system to compute the reputation value. When a UAV offloads a task, its reputation is debited from its total reputation. On the other hand, the reputation is credited to that UAV to which the task is offloaded. We apply a reputation transfer-based bargaining game while prioritizing the UAVs with higher capability. In particular, we adopt <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Rubinstein bargaining</i> model to evaluate the trust of the UAVs and design an incentivized offloading mechanism. We use the concept of <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Subgame Perfect Nash Equilibrium</i> (<italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">SPNE</i>) to decide an optimal amount of reputation to be transferred between two UAVs. The game is modeled in a way that encourages UAVs with lower reputations to participate in the game by making them more patient so that they can negotiate for higher fractions of reputation. Simulation results show that TOFU reduces delay and energy consumption per task by up to 1.74% and 2%, respectively. Additionally, TOFU reduces task failure rate by approximately 9% as compared to the traditional approaches such as DeTTO, DPTO, and SOA, respectively.