Improving Knowledge Based Detection of Soft Attacks Against Autonomous Vehicles with Reputation, Trust and Data Quality Service Models
Autonomous vehicles group’s security and safety improvement and assurance is a challenging research problem. In this paper, we describe our smart data-oriented security service, which is aimed at detecting malfunctioning or malicious agents based on the fusion of multi-agents Reputation, Trust and Data Quality (DQ) models for traffic control. To address the classical Reputation zero value challenge, we introduce the DQ evaluation service, which allows to use the vehicle’s objective characteristics to assign the initial Reputation value to a new agent when it is joining the group. To validate our approach, we conducted an empirical study on real intersection traffic with multiple vehicles. Multiple experiments were performed on our custom physical intersection management test ground and even bigger vehicles groups were studied by simulation. The experimental results verify our approach capability to effectively detect malfunctioning and malicious agents. The empirical study confirmed that the DQ service improves detection performance.
Chuprov, S., Viksnin, I., Kim, I., Melnikov, T., Reznik, L., & Khokhlov, I. (2021, September 5). Improving knowledge based detection of soft attacks against autonomous vehicles with reputation, trust and data quality service models [conferrence presentation]. 2021 IEEE International Conference on Smart Data Services (SMDS), Chicago, IL. Doi: 10.1109/SMDS53860.2021.00025