SECAUCTEE: Securing Auction Smart Contracts using Trusted Execution Environments

Abstract

Smart contracts running on blockchains have emerged as an indispensable mechanism to enhance trust, security, transparency and traceability of data shared among critical distributed applications. Unfortunately, a smart contract deployed on a blockchain by itself is usually inadequate in maintaining data security and privacy because the data are replicated to all the nodes on the network. There has been some recent work that tries to tackle this privacy leakage issue in smart contract execution by integrating blockchains with hardware supported trusted execution environments(TEEs). Although TEEs ensure privacy to some extent, the smart contract execution can still be compromised if the developed code does not use the TEEs' capabilities correctly. One important security issue for leveraging TEEs in practice is the memory access pattern disclosure. Even though the TEEs encrypt all the memory content during the program execution, the memory access sequence can be observed by the malicious operating system, and can be used to infer sensitive information such as “who submitted the second highest bid to the auction?”. Hence, for enhanced security for TEE based applications, the memory access pattern leakage need to be addressed. Given these observations, an apparent question that comes to light is, how can we use TEEs correctly to enable efficient, privacy enhancing and secure applications? In this work, we address this challenge in the context of digital auctions. We develop a novel generic and secure framework that allow an auction smart contract to run inside secure enclaves over Intel SGX based TEEs on a blockchain. To our knowledge, this is the first work that provides access pattern leakage free TEE based secure auction smart contract deployment. We achieve this by implementing oblivious execution (i.e., no memory access pattern leakage) of both first price and second price sealed bid auctions as templates. Furthermore, we implement an end-to-end encryption service to keep the bids secure. Our empirical results and privacy analysis show that this architecture does not cause a significant impact to efficiency given the level of security achieved.