Abstract: As the fifth-generation (5G) mobile communication technology features high speed, low latency, and wide coverage, it has been widely applied in vehicular ad hoc networks (VANETs) to improve information transmission efficiency. However, in the context of vehicular networks, communication occurs in a highly open environment, and transmitted information is susceptible to tampering, interception, and deletion by malicious attackers. Therefore, the security of vehicular communication needs to be ensured, especially in terms of information authentication and integrity. In addition, vehicles need to protect their identity privacy during communication to avoid malicious attackers from tracking their driving routes. At the same time, if vehicles engage in malicious behavior, the Trusted Authority (TA) need to be able to track malicious vehicles and reveal their true identities. Although Group Signature (GS) can achieve anonymity and traceability, its security relies on a strong assumption that the secret key is stored on the device for a long time and is not accessible to any malicious attackers. However, in reality, attackers can steal the secret keys stored on devices through physical attacks similar to side-channel attacks, which greatly threatens communication security. To address this issue, this study proposes a anonymous authentication protocol resistant to physical theft for 5G VANETs. The protocol uses Physically Unclonable Function (PUF) to construct a novel group signature algorithm for anonymous authentication between Vehicle-to-Vehicle (V2V) and Vehicle-to-Infrastructure (V2I). The Challenge-Response Pair (CRP) of PUF is used to generate keys for vehicles in real-time, without the need for long-term storage of keys on vehicles, in order to avoid physical theft by malicious attackers. In addition, when disputes arise, the TA can track malicious vehicles through signatures. Finally, security analysis shows that the protocol meets security requirements such as resistance to physical theft attacks, unforgeability, anonymity, traceability, non-repudiation, and unlinkability. Furthermore, performance evaluation shows that compared with other relevant protocols, this protocol performs better in terms of computational and communication costs.