fate隐私求交案例

intersect组件是解决纵向联邦学习中的隐私求交问题

fate隐私求交的方式有三种：raw,rsa,dh。raw方式不安全，rsa和dh方式是安全的，dh是基于对称加密的安全交集 rsa是基于RSA(非对称加密)的安全交集，，dh方法也用于安全的信息检索（SIR），fate intersect 支持多主机模式即(1个guest与多个host进行求交集)

可配置的hash方法有sha256、md5 和 sm3。raw交集支持base64编码，支持与缓存的交集。

统案例脚本文件：/data/projects/fate/examples/dsl/v2/

1、 1v1案例：

本次案例在guest执行任务，求交host

/data/projects/fate/examples 目录创建测试目录，并拷贝系统配置

cd /data/projects/fate/examples/mytest cp /data/projects/fate/examples/dsl/v2/intersect/test_intersect_job_dsl.json ./ cp /data/projects/fate/examples/dsl/v2/intersect/test_intersect_job_raw_conf.json ./

1.1上传文件

1.1.1、host创建上传脚本：

upload_xxl_host.json

{ "file": "/data/projects/fate/examples/mytest/xxl_test_host.csv", "head": 1, "partition": 10, "work_mode": 0, "namespace": "sp_host", "table_name": "tb_host" }

xxl_test_host.csv 是数据文件（需要有表头）

namespace 表空间名称

table_name 表名称

1.1.2、guest创建脚本：

upload_xxl_guest.json

{ "file": "/data/projects/fate/examples/mytest/xxl_test_guest.csv", "head": 1, "partition": 10, "work_mode": 0, "namespace": "sp_guest", "table_name": "tb_guest" }

1.1.3上传文件

# host 端：
source /data/projects/fate/bin/init_env.sh
flow data upload -c upload_xxl_host.json

# guest端：
source /data/projects/fate/bin/init_env.sh
flow data upload -c upload_xxl_guest.json

1.2创建任务脚本

1.2.1、创建dsl文件

test_intersect_job_dsl.json

{ "components": { "reader_0": { "module": "Reader", "output": { "data": [ "data" ] } }, "data_transform_0": { "module": "DataTransform", "input": { "data": { "data": [ "reader_0.data" ] } }, "output": { "data": [ "data" ], "model": [ "model" ] } }, "intersect_0": { "module": "Intersection", "input": { "data": { "data": [ "data_transform_0.data" ] } }, "output": { "data": [ "data" ] } } } }

1.2.2、创建任务配置文件

test_intersect_job_rsa_conf.json

{ "dsl_version": 2, "initiator": { "role": "guest", "party_id": 9999 }, "role": { "guest": [ 9999 ], "host": [ 10000 ] }, "component_parameters": { "common": { "intersect_0": { "intersect_method": "rsa", "sync_intersect_ids": false, "only_output_key": true, "rsa_params": { "hash_method": "sha256", "final_hash_method": "sha256", "split_calculation": false, "key_length": 2048 } } }, "role": { "guest": { "0": { "reader_0": { "table": { "name": "tb_guest", "namespace": "sp_guest" } }, "data_transform_0": { "with_label": false, "output_format": "dense" } } }, "host": { "0": { "reader_0": { "table": { "name": "tb_host", "namespace": "sp_host" } }, "data_transform_0": { "with_label": false, "output_format": "dense" } } } } } }

1.3执行任务（guest端执行）

cd /data/projects/fate/examples/mytest
source /data/projects/fate/bin/init_env.sh
flow job submit -d test_intersect_job_dsl.json -c test_intersect_job_raw_conf.json

2、1v2案例

本次案例在guest上执行求交任务，求交host2个文件

2.1 上传文件

2.1.1 host创建上传脚本（2个）：

upload_xxl_host.json

# 上传第一个文件：upload_xxl_host1.json { "file": "/data/projects/fate/examples/mytest/1v2/xxl_test_host1.csv", "head": 1, "partition": 10, "work_mode": 0, "namespace": "sp_host1", "table_name": "tb_host1" } # 上传第二个文件：upload_xxl_host2.json { "file": "/data/projects/fate/examples/mytest/xxl_test_host2.csv", "head": 1, "partition": 10, "work_mode": 0, "namespace": "sp_host2", "table_name": "tb_host2" }

2.1.2 guest创建脚本：

upload_xxl_guest.json

{ "file": "/data/projects/fate/examples/mytest/1v2/xxl_test_guest.csv", "head": 1, "partition": 10, "work_mode": 0, "namespace": "sp_guest1", "table_name": "tb_guest1" }

2.1.3 上传文件

# host 端：
source /data/projects/fate/bin/init_env.sh
cd /data/projects/fate/examples/mytest/1v2

flow data upload -c upload_xxl_host1.json

flow data upload -c upload_xxl_host2.json

# guest端：
source /data/projects/fate/bin/init_env.sh
cd /data/projects/fate/examples/mytest/1v2

flow data upload -c upload_xxl_guest.json

2.2 创建任务脚本

本次任务在guest上执行，任务脚本在guest端

2.2.1 创建dsl文件

test_union_job_dsl.json

{ "components": { "reader_0": { "module": "Reader", "output": { "data": [ "data" ] } }, "reader_1": { "module": "Reader", "output": { "data": [ "data" ] } }, "intersection_0": { "module": "Intersection", "input": { "data": { "data": [ "reader_0.data" ] } }, "output": { "data": [ "data" ] } }, "intersection_1": { "module": "Intersection", "input": { "data": { "data": [ "reader_1.data" ] } }, "output": { "data": [ "data" ] } }, "union_0": { "module": "Union", "input": { "data": { "data": [ "intersection_0.data", "intersection_1.data" ] } }, "output": { "data": [ "data" ] } } } }

2.2.2 创建任务配置文件

test_union_job_conf.json

注意：因为 guest 只有1个文件，host 有2个文件，guest 1个文件求交 host 2个文件，所以这里 guest 角色的 reader_ 和 reader_1 读取的数据都是 guest 同一个表的同一份数据

{ "dsl_version": 2, "initiator": { "role": "guest", "party_id": 9999 }, "role": { "host": [ 10000 ], "guest": [ 9999 ] }, "component_parameters": { "common": { "union_0": { "allow_missing": false, "need_run": true } }, "role": { "guest": { "0": { "reader_0": { "table": { "name": "tb_guest1", "namespace": "sp_guest1" } }, "reader_1": { "table": { "name": "tb_guest1", "namespace": "sp_guest1" } } } }, "host": { "0": { "reader_0": { "table": { "name": "tb_host1", "namespace": "sp_host1" } }, "reader_1": { "table": { "name": "tb_host2", "namespace": "sp_host2" } } } } } } }

2.3 执行任务（guest端执行）

cd /data/projects/fate/examples/mytest/1v2
source /data/projects/fate/bin/init_env.sh

flow job submit -d test_intersect_job_dsl.json -c test_intersect_job_raw_conf.json

正确求交页面显示：