Research on State-Based Bluetooth Multi-Protocol Fuzzing

As Bluetooth technology evolves, security issues highlight the need for ongoing attention amid progress and updates. Research on Bluetooth protocol security is becoming increasingly critical. Presently, most research relies on manual construction methods to generate state machine models. However, manual construction is error-prone, time-consuming and single type of test. In this paper, we contribute that: (1) Inserting an interface function at the Bluetooth controller layer achieves data packet modification, enabling seamless extension of upper-layer protocol stacks. (2) Adopting an automated machine learning approach to learn the state transition relationships of Bluetooth protocols, yielding the protocol’s state transition relations. (3) Discovering eight distinct Bluetooth vulnerabilities across six Bluetooth devices by our method. These vulnerabilities comprise three known and five unknown vulnerabilities. The experimental results validate the effectiveness of the proposed method.