Change Log
V1.1.0 Mar 2022
Features
- dai
- model evaluation: support model evaluation and live evaluation through random split, cross validation or LOO
- self-computing: support two computing modes for executors, one is to compute using others' data, one is to compute using their own data
Improvements
- crypto
- improve security of pairing-based PoRH: upgrade algorithm to avoid mod-N attack, which enables storage nodes to pass challenges using only slice data mod N
V1.0.0 Jan 2022
This is the first public release of PaddleDTX.
Features
-
dai
- vertical federated learning(linear regression and logistic regression)
- training/prediction task management
- p2p
- requester cli
- support xuperchain
-
xdb
- file upload and download
- file slicing and encryption
- file copies making and distribution
- proof of replication holding(pairing/merkle)
- health detection
- file migration
- resource access control
- support xuperchain and fabric
-
crypto
- distributed implementation of linear regression
- distributed implementation of logistic regression
- additive homomorphic encryption(Paillier)
- private set intersection
- oblivious transfer
- secret sharing
- proof of replication holding