|
|
@@ -8,22 +8,36 @@ CONTENTS OF THIS FILE
|
|
|
|
|
|
INTRODUCTION
|
|
|
------------
|
|
|
+ `/preprocessing` holds code for the 3-Party preprocessing phase
|
|
|
+ `/2p-preprocessing` holds the OT-based code 2-Party preprocessing phase
|
|
|
+ `/duoram-online` holds the code for online phase of the 3-Party Duoram and 2-Party reads
|
|
|
+ `/cpir-read` includes the code CPIR-based reads
|
|
|
|
|
|
|
|
|
Setting up the docker
|
|
|
------------
|
|
|
-
|
|
|
+- `sh build-docker.sh` builds the docker image
|
|
|
+- `sh run-docker.sh` launches three docker containers for parties P2, P0, and P1
|
|
|
+- `sh set-networking.sh` sets up the latencies and bandwidth constraints among the three docker containers launched; pass arguments of latency and bandwidth
|
|
|
+- For example `sh set-networking.sh 1ms 40mbit`.
|
|
|
+- `sh unset-networking.sh` removes the latency and bandwidth settings.
|
|
|
+- `sh experiments.sh` runs the pre-processing phase of the 3-party computation.
|
|
|
+- `sh experiments-online.sh` runs the online phase of the 3-party computation.
|
|
|
+- `sh experiments-OT.sh` runs the OTs to generate blinds of 2-party preprocesing phase.
|
|
|
+- `sh experiments-2p-preproc.sh` runs the pre-processing phase of the 2-party computation.
|
|
|
+- `sh experiments-cpir.sh` runs the CPIR-based reads.
|
|
|
+- `sh cleanup.sh` stops the three dockers.
|
|
|
|
|
|
Running the code
|
|
|
-------------
|
|
|
|
|
|
1. Prepreprocessing:
|
|
|
- Let ip0, ip1, and ip2 be the IP addresses of the three docker containers. Consider a database of size `2^N`. To generate DPFs (with `t` threads) to do `w` writes and `r` reads.
|
|
|
+ Let ip0, ip1, and ip2 be the IP addresses of the three docker containers. Consider a database of size `2^N`. Suppose the goal is to generate `r` DPFs.
|
|
|
The pre-processing generates DPFs with leaves of 128 bits. We use the first 64-bits for the writes. The remaining 64-bits are used for share conversion (XOR shared-flag vectors to additive shares)
|
|
|
- 1.1 `cd duoram/preprocessing-aes`
|
|
|
- 1.2 `./p2preprocessing ip1 ip0 t N w r`
|
|
|
- 1.3 `./preprocessing1 ip2 ip0 t N w r`
|
|
|
- 1.4 `./preprocessing0 ip2 ip1 t N w r`
|
|
|
+ 1.1 `cd duoram/preprocessing`
|
|
|
+ 1.2 `./p2preprocessing ip1 ip0 N r`
|
|
|
+ 1.3 `./preprocessing1 ip2 ip0 N r`
|
|
|
+ 1.4 `./preprocessing0 ip2 ip1 N r`
|
|
|
|
|
|
This writes the evaluations of DPFs into a file which we use in the online phase.
|
|
|
|
|
|
@@ -31,10 +45,12 @@ Running the code
|
|
|
Let ip0, ip1, and ip2 be the IP addresses of the three docker containers. Consider a database of size `2^N`.
|
|
|
2.1 `cd duoram/duoram-online`
|
|
|
2.2 `./p2 ip1 ip0 N W D I A`
|
|
|
- 2.3 `./duoram1 ip1 ip0 N W D I A`
|
|
|
- 2.4 `./duoram0 ip1 ip0 N W D I A`
|
|
|
+ 2.3 `./duoram1 ip0 ip2 N W D I A`
|
|
|
+ 2.4 `./duoram0 ip1 ip2 N W D I A`
|
|
|
|
|
|
-
|
|
|
+ Important Warning
|
|
|
+ --------------------
|
|
|
+This software is for performance testing ONLY! The purpose of this software is to evaluate the performance of the system, NOT to be used in a deployment scenario.
|
|
|
|
|
|
|
|
|
|
avadapal