多方安全计算（MPC）任务
多方安全计算（Multi-Party Computation，MPC）功能是指多个参与方在不暴露私有数据的情况下，共同完成一项计算任务的过程。
MPC任务
MPC LR（逻辑回归）
./primihub-cli --task_config_file="example/mpc_lr_task_conf.json"
MPC 四则运算
./primihub-cli --task_config_file="example/mpc_add_task_conf.json"
./primihub-cli --task_config_file="example/mpc_sub_task_conf.json"
./primihub-cli --task_config_file="example/mpc_mul_task_conf.json"
./primihub-cli --task_config_file="example/mpc_div_task_conf.json"
MPC 联合统计
对于联合统计，当前平台已经实现了最大值，最小值，平均值，以及求和的功能
提交统计最大值任务
./primihub-cli --task_config_file="example/mpc_statistics_max_task_conf.json"
提交统计最小值任务
./primihub-cli --task_config_file="example/mpc_statistics_min_task_conf.json"
提交统计平均值任务
./primihub-cli --task_config_file="example/mpc_statistics_avg_task_conf.json"
提交统计总和任务
./primihub-cli --task_config_file="example/mpc_statistics_sum_task_conf.json"
MPC 数据预处理任务
对于使用的数据集，有时候并不是完全就可以直接使用的，在数据采集过程中会有部分数据缺失，异常以及其他的问题存在，因此，对数据进行预处理是必不可少的步骤，在隐私计算中同样适用，在通过数据使用隐私计算平台的过程中也需要根据数据的真实情况对数据做部分预处理。 对于横向场景，实现了缺失值/异常值的处理流程，可通过如下命令发起相关任务
./primihub-cli --task_config_file="example/mpc_missing_value_task_conf.json"
MPC LR 任务示例详解
提交MPC LR任务
执行一个MPC的逻辑回归任务：
如果是通过下载二进制文件或本地编译启动，编译完成后在代码根目录下执行以下命令；如果是通过docker-compose启动，先执行 docker exec -it primihub-node0 bash 进入到 primihub-node0 容器中，再执行以下命令。
./primihub-cli --task_config_file="example/mpc_lr_task_conf.json"
分别观察node0、node1和node2的日志，有结果文件生成则代表任务运行成功。
node0:
···
I20230614 17:46:15.512172 48 node.cc:749] start to schedule task, task_type: 0
I20230614 17:46:15.512256 48 node.cc:1069] 🤖️ Start create worker node0 worker id: 28ca8b00-c1bf-414d-a9f3-0752d2f1a55c
I20230614 17:46:15.512315 48 node.cc:1073] 🤖️ Fininsh create worker node0 worker id: 28ca8b00-c1bf-414d-a9f3-0752d2f1a55c
E20230614 17:46:15.512459 48 proto_parser.cc:56] party_datasets: 3
W20230614 17:46:15.541919 48 parser.cc:192] update access info for party_name: PARTY0
W20230614 17:46:15.542023 48 parser.cc:192] update access info for party_name: PARTY1
W20230614 17:46:15.542083 48 parser.cc:192] update access info for party_name: PARTY2
I20230614 17:46:15.542430 48 aby3_scheduler.cc:191] Dispatch SubmitTask to 3 node
I20230614 17:46:15.542496 48 aby3_scheduler.cc:203] Dispatch Task to party: node1:primihub-node1:50051:0: party_name: PARTY1
I20230614 17:46:15.542651 48 aby3_scheduler.cc:203] Dispatch Task to party: node2:primihub-node2:50052:0: party_name: PARTY2
I20230614 17:46:15.542924 48 aby3_scheduler.cc:203] Dispatch Task to party: node0:primihub-node0:50050:0: party_name: PARTY0
I20230614 17:46:15.542958 62 aby3_scheduler.cc:49] dest node node1:primihub-node1:50051:0:
I20230614 17:46:15.543244 63 aby3_scheduler.cc:49] dest node node2:primihub-node2:50052:0:
I20230614 17:46:15.543543 64 aby3_scheduler.cc:49] dest node node0:primihub-node0:50050:0:
I20230614 17:46:15.549181 61 node.cc:848] start to create worker for task: job_id : 100 task_id: 200 request id: 28ca8b00-c1bf-414d-a9f3-0752d2f1a55c
I20230614 17:46:15.549252 61 node.cc:1069] 🤖️ Start create worker node0 worker id: 28ca8b00-c1bf-414d-a9f3-0752d2f1a55c
I20230614 17:46:15.549294 61 node.cc:1073] 🤖️ Fininsh create worker node0 worker id: 28ca8b00-c1bf-414d-a9f3-0752d2f1a55c
I20230614 17:46:15.549533 61 node.cc:858] create worker thread future for task: job_id : 100 task_id: 200 request id: 28ca8b00-c1bf-414d-a9f3-0752d2f1a55c finished
I20230614 17:46:15.549640 67 node.cc:821] begin to execute task
I20230614 17:46:15.552394 66 worker.cc:122] Request id: 28ca8b00-c1bf-414d-a9f3-0752d2f1a55c update party: PARTY0 status to: RUNNING
I20230614 17:46:15.553223 67 worker.cc:70] Worker start execute task
I20230614 17:46:15.553269 67 logistic.cc:131] party_name: PARTY0
I20230614 17:46:15.553288 61 node.cc:875] create worker thread for task: job_id : 100 task_id: 200 request id: 28ca8b00-c1bf-414d-a9f3-0752d2f1a55c finished
I20230614 17:46:15.553335 67 logistic.cc:158] Train data train_party_0, test data .
I20230614 17:46:15.554432 68 worker.cc:122] Request id: 28ca8b00-c1bf-414d-a9f3-0752d2f1a55c update party: PARTY2 status to: RUNNING
I20230614 17:46:15.554865 68 worker.cc:122] Request id: 28ca8b00-c1bf-414d-a9f3-0752d2f1a55c update party: PARTY1 status to: RUNNING
I20230614 17:46:15.565569 67 logistic.cc:430] local_id_local_id_: 2
I20230614 17:46:15.565621 67 logistic.cc:431] next_party: PARTY1 detail: node1:primihub-node1:50051:0:
I20230614 17:46:15.565647 67 logistic.cc:432] prev_party: PARTY2 detail: node2:primihub-node2:50052:0:
I20230614 17:46:15.565676 67 mpc_channel.cc:11] Init channel to PARTY1.
I20230614 17:46:15.565696 67 mpc_channel.cc:11] Init channel to PARTY2.
I20230614 17:46:15.565726 67 runtime.h:48] Init Runtime : 2, 0x7f2a5000a060
I20230614 17:46:15.565752 67 aby3ML.cc:35] Runtime init finish.
I20230614 17:46:15.606828 67 aby3ML.cc:39] Encryptor init finish.
I20230614 17:46:15.624158 67 aby3ML.cc:42] Evaluator init finish.
E20230614 17:46:15.624224 67 logistic.cc:462] Construct shares of train data and train label.
E20230614 17:46:15.718981 67 logistic.cc:470] end Construct shares of train data and train label.
I20230614 17:46:15.757925 67 logistic.cc:584] Train dataset has 558 examples, dimension of each is 11.
I20230614 17:46:15.757994 67 logistic.cc:587] Test dataset has 141 examples, dimension of each is 11.
I20230614 17:46:15.758018 67 logistic.cc:49] (Learning rate):0.0078125.
I20230614 17:46:15.758060 67 logistic.cc:50] (Epoch):100.
I20230614 17:46:15.758076 67 logistic.cc:51] (Batchsize) :128.
I20230614 17:46:15.758090 67 logistic.cc:52] (Train_loader size):4.
I20230614 17:46:15.976900 67 regression.h:249] 0 @ percent:0.326241.
I20230614 17:46:17.106693 67 regression.h:249] 10 @ percent:0.893617.
I20230614 17:46:18.369973 67 regression.h:249] 20 @ percent:0.921986.
I20230614 17:46:19.758834 67 regression.h:249] 30 @ percent:0.929078.
I20230614 17:46:20.961714 67 regression.h:249] 40 @ percent:0.929078.
I20230614 17:46:22.073453 67 regression.h:249] 50 @ percent:0.929078.
I20230614 17:46:23.093506 67 regression.h:249] 60 @ percent:0.929078.
I20230614 17:46:24.239471 67 regression.h:249] 70 @ percent:0.929078.
I20230614 17:46:25.444774 67 regression.h:249] 80 @ percent:0.929078.
I20230614 17:46:26.469460 67 regression.h:249] 90 @ percent:0.93617.
I20230614 17:46:27.379612 67 logistic.cc:607] Party 2 train finish.
I20230614 17:46:27.384567 67 logistic.cc:667] Save model to data/result/mpc_lr_model.csv.
I20230614 17:46:27.420114 3032 worker.cc:119] collected finished party count: 1
I20230614 17:46:27.420164 3032 worker.cc:122] Request id: 28ca8b00-c1bf-414d-a9f3-0752d2f1a55c update party: PARTY2 status to: SUCCESS
I20230614 17:46:27.424995 3032 worker.cc:119] collected finished party count: 2
I20230614 17:46:27.425034 3032 worker.cc:122] Request id: 28ca8b00-c1bf-414d-a9f3-0752d2f1a55c update party: PARTY1 status to: SUCCESS
I20230614 17:46:27.428491 67 logistic.cc:673] Register new dataset finish.
I20230614 17:46:27.429445 3035 worker.cc:119] collected finished party count: 3
I20230614 17:46:27.429504 3035 worker.cc:122] Request id: 28ca8b00-c1bf-414d-a9f3-0752d2f1a55c update party: PARTY0 status to: SUCCESS
node1 和 node2 日志和 node0 类似，省略。
MPC任务除了支持逻辑回归外、还支持四则运算、求均求和最大最小值统计、缺失值异常值处理等，具体各任务的示例配置请参考 example目录下的mpc_*.json 文件
LR任务参数说明
| 参数 | 数据类型 | 参数示例 | 参数说明 |
|---|---|---|---|
| task_code.code | logistic_regression | MPC的任务类型 | |
| params.BatchSize | INT32 | 128 | 数据大小 |
| params.NumIters | INT32 | 100 | 迭代次数 |
| params.modelName | STRING | data/result/mpc_lr_mode.csv | 生成模型的存储路径及文件名 |
| party_datasets.PARTY0 | STRING | test_party_0,train_party_0 | 训练数据集 |
| party_datasets.PARTY1 | STRING | test_party_1,train_party_1 | 训练数据集 |
| party_datasets.PARTY2 | STRING | test_party_2,train_party_2 | 训练数据集 |