ZeroTrace(ZT) is the first system that enabled instantiations of Oblivious-RAMs(ORAMs) on a server-device that supports Intel-SGX. ZT is also secure against known side-channel attacks against SGX.* Oblivous RAM or ORAMs have been theoretically known for a long while now [1], with the advent of secure hardware modules like Intel SGX, we notice an opportunity to make these incredible cryptographic primitives, deployable in practical scenarios.
Sajin Sasy, Sergey Gorbunov, and Christopher Fletcher. "ZeroTrace: Oblivious memory primitives from Intel SGX." Symposium on Network and Distributed System Security (NDSS). 2018.
ZeroTrace employs an oblivious variant of traditional ORAM controller logic. To do so within an SGX environment, without being susceptible to side-channel attacks, the only truly TRUSTED space that we can use are the CPU registers. Hence function snippets that require conditional operations are rewritten to be oblivious via linear scans and assembly functions that leverage CMOV instructions wherever appropriate.
Currently ZeroTrace supports two ORAM backends, PathORAM [2] and CircuitORAM [3]. Please refer the corresponding papers to get a more concrete understanding of how to parameterize these ORAMs for secure deployment scenarios.
ZeroTrace requires a fully functional Intel SGX-SDK stack, we tested it with SGX SDK Linux 2.1 release. To set up your SGX-SDK stack, please refer to: https://github.com/intel/linux-sgx
ZeroTrace was tested with the following configuration :
- 64-bit Ubuntu 16.04 (with 4.4.0 kernel)
- gcc = 5.4.0
- nasm = 2.11.08 (Almost all 2.+ versions of nasm should be sufficient)
ZeroTrace takes a LOT of command line parameters. This was done in the interest of making it easy to instantiate ORAMs with different parameters to test performance across a wider range of possible parameters. In order to make it simpler to use, a shell script "exec_zt.sh" is provided, in which one can more conveniently customize the parameters to run ZeroTrace with.
To build the ZeroTrace library: In the ZeroTrace folder, execute:
make clean
make
This will produce the ZeroTrace library (libZT.so) and copy it to the Sample_App folder. To execute the Sample_Application provided by us, simply execute (Change the parameters in the zt_exec.sh script as needed by your application)
./exec_zt.sh
The Sample_App is a simple example code that shows how to integrate ZT with native C/C++ code. But it should be sufficient in demonstrating how to use this. Do let us know if we can help you in anyway integrate it with your application/product.
ZT_Initialize() : An initialization function that sets up the ZeroTrace Enclave, before one can create and access ORAMs through it.
ZT_New(args) : This function creates a new ORAM instance with the provided parameters. Currently all data blocks of these instantiated trees are dummy data, to populate the tree one has to perform individual writes to the tree after instantiating it. ZeroTrace supports multiple ORAM instances, this function returns an instance_id which is used to keep track of the ORAM instance of this newly created ORAM, so that the user can query this particular ORAM instance later.
ZT_Access(args) : This function is used to access (read/write) values to a previously-created ORAM Tree.
ZT_Bulk_Read(args) : This is a newly supported function that enables users to do bulk reads, without performing enclave exits and entry for each request individually.
The file ZT.hpp can be used as a reference for the underlying arguments, the argument names are self-explanatory.
-
ZeroTrace assumes the enclave and client has already performed a Remote Attestation handshake and established a shared secret key. ZeroTrace was designed to be used as a framework for research, hence it uses a hardcoded key (as this shared secret key) and IV as you will notice from the source. It is easy to replace them with genuine key sampling functions (which in most cases are already present in the source, but just hijacked with static values to make it easy to debug and experiment).
-
ZeroTrace was designed to be a framework for experimenting with different ORAMs, in this intersection of secure hardware and ORAMs. It is my hope that we will see other contributors use this tool to either develop ORAM backends for other known ORAM designs, or possibly even design their own ORAM schemes and test it out using ZeroTrace. You will notice that the class ORAMTree, provides a sufficient abstraction for rapid-deployment of almost any Tree-based ORAM scheme.
-
Currently, the HDD backend code is broken, so is the Store/Resume functionality. Hence applications must use "memory" as the backend, and "new" for the new/resume flag in the command line parameters/zt_exec.sh script. (I will be uploading patches to stabilize that soon.)
-
An integrations with Eleos, is still pending, and on the TO-DO list, to bump performance up a bit more.
Feel free to reach out to me to get help in setting up our system or any other queries you may have related to ZeroTrace: sshsshy7@gmail.com
1 - Goldreich, O. and Ostrovsky, R., 1996. Software protection and simulation on oblivious RAMs. Journal of the ACM (JACM), 43(3), pp.431-473.
2 - Stefanov, E., Van Dijk, M., Shi, E., Fletcher, C., Ren, L., Yu, X. and Devadas, S., 2013, November. Path ORAM: an extremely simple oblivious RAM protocol. In Proceedings of the 2013 ACM SIGSAC conference on Computer & communications security (pp. 299-310). ACM.
3 - Wang, X., Chan, H. and Shi, E., 2015, October. Circuit oram: On tightness of the goldreich-ostrovsky lower bound. In Proceedings of the 22nd ACM SIGSAC Conference on Computer and Communications Security (pp. 850-861). ACM.
* Recently we have seen attacks against the Intel special service enclaves, that can compromise the enclave key generation process. We expect Intel to patch these bugs, as these effectively cripple SGX as a whole.