Adrien Ghosn, Prof. James R. Larus, Prof. Edouard Bugnion.
This repository is a snapshot of GOTEE, a research prototype compiler based on the Go programming language. The corresponding research paper was accepted at ATC19.
Trusted Execution Environments (TEEs), such as Intel SGX enclaves, use hardware to ensure the confidentiality and integrity of operations on sensitive data. While the technology is available on many processors, the complexity of its programming model and its performance overhead have limited adoption. TEEs provide a new and valuable hardware functionality that has no obvious analogue in programming languages, which means that developers must manually partition their application into trusted and untrusted components.
This paper describes an approach that fully integrates trusted execution into a language. We extend the Go language to allow a programmer to execute a goroutine within an enclave, to use low-overhead channels to communicate between the trusted and untrusted environments, and to rely on a compiler to automatically extract the secure code and data. Our prototype compiler and runtime, GOTEE , is a backward-compatible fork of the Go compiler.
The evaluation shows that our compiler-driven code and data partitioning efficiently executes both microbenchmarks and applications. On the former, GOTEE achieves a 5.2× throughput and a 2.3× latency improvement over the Intel SGX SDK. Our case studies, a Go ssh server, the Go tls package, and a secured keystore inspired by the go-ethereum project, demonstrate that minor source-code modifications suffice to provide confidentiality and integrity guarantees with only moderate performance overheads.
GOTEE is a research prototype, developed by Adrien Ghosn as part of his PhD at EPFL (Switzerland). The code is not production-ready and comes with no guarantees. Furthermore, the original repository for GOTEE is an ongoing project. As a result, APIs, semantics, and features might change. We will update this snapshot everytime a milestone is reached.
Gotee only supports linux x86 platforms.
Gotee has one dependency on a serializer, required to communicate with the Intel AESM module. This package must be installed and the serializer command available in your shell environment (go install command).
Moreover, you need to have an SGX enabled processor with the Intel AESM enclave running (for more information about how to set up Intel SGX, go to Intel SGX SDK repositories. The Intel SGX SDK itself is not required.
Gotee requires a valid go install on the target machine to bootstrap the compilation.
To compile Gotee, simply use make in the root directory of this folder.
This will trigger the bootstrap compilation and install Gotee as a system command called gotee.
The make clean command allows to uninstall gotee.
The scripts try to install Gotee under /usr/local/bin using a symbolic link.
You might have to change access rights to this folder in order to let the script install the command.
A hello-world example is included in this repository under example/hello-world.
A detailed README is available in this folder.
Due to a late refactoring and the prototype nature of the compiler, Gotee temporarily prevents using gosecure on functions declared in the main package, and discourages using the gosecure keyword outside of the main package.
We further require the gosec package to be explicitly imported in main.
This restrictive behavior was introduced during a code-refactoring that aimed at cleaning-up and improving the implementation, and will be fixed. For the moment, the compiler emits errors if it detects a violation.
A simple workaround is to provide a public function for a particular package that needs to trigger secured routines, and call this function with gosecure from main:
package securedpkg
func Spawn() {
go secured1()
go secured2()
...
}
...
package main
gosecure securedpkg.Spawn()
Instead of:
package securedpkg
gosecure secured1()
gosecure secured2()
...
These restrictions were introduced as a temporary fix to prevent nested gosecure calls. A cleaner approach is being developped.
We are working on cleaning the benchmarks used in the paper and making them open-source as well.
https://github.com/epfl-dcsl/gosecure-benchmarks
HEAD master on 05.29.2019