Research On GPU-based Convert Channel

GPUs play an increasingly important role in compute-intensive and mission-critical applications,especially in cryptography,finance,and defense.As an accelerator,GPUs can speed up the running of application software and are currently being used in a large number of applications.As the performance of GPUs continues to increase and the range of applications continues to increase,GPUs are beginning to receive attention from attackers.However,the security of GPUs has not been thoroughly studied.Some attack methods for the CPU are also suitable for GPUs,such as covert channel attacks.Covert channel can make a malicious application without network access transfer data to another application to leak device data.At the same time,covert channels can bypass the protection of sensitive information exposure,such as sandboxing or information flow tracking,resulting in sensitive data fleeing containment.Therefore,attacking GPUs with covert channels can reveal confidential information stored in GPUs.Currently covert channels have been implemented on GPUs constant buffers and special functional units.However,the covert channel attack method based on other memory structures of GPUs can also be implemented,and even has higher threat.Therefore,this paper mainly studies hidden channel attacks.The main work of this thesis is as follows:(1)This paper proposes to implement covert channels on GPUs global memory.Global memory has high bandwidth and requires atomic operations to observe significant differences in access time.Global memory can be accessed by all threads,making it easier to leak confidential information,which is more harmful than previous GPUs constant cache covert channels and special function covert channels.(2)This paper designs a synchronous data transmission mechanism to ensure that two functions perform synchronous data transmission.The information transfer on the covert channel is not all reliable.Because the high bandwidth of the global memory may cause noise to affect the data transmission,the covert channel loses synchronization,which affects the correctness of the transmitted data.Therefore,a synchronous data transmission mechanism is proposed,which ensures the reliability of covert channel data transmission.(3)This article successfully implements covert channel attacks.Perform detailed analysis of covert channel transmissions by using covert channels for pictures,text files,and audio files,and in ideal environments and noisy environments.Aiming at the covert channel transmission process under noisy environment,a method of data correction using Hamming code is proposed.