v1.1.4-alpha

Added

• [important] Now the Java implementation of SEAL is fully compatible with SEAL 4.0.0. That is, one can use SEAL to serialize any parameters then use our Java implementation to deserialize, and verse visa.
• [important] We implement CKKS in the Java implementation of SEAL.
• [important] We merge the implementation of our paper SCAPE into mpc4j (See submodule mpc4j-work-scape).

v1.1.3 (beta)

This version is for Artifact Evaluation of our paper "Practical Keyword Private Information Retrieval from Key-to-Index Mappings" accepted to USENIX Security 2025.

[1.1.3]

Added

• [important] We update rpc implementations and now we do not further request starting parties in sequence.
• We implement FrodoPIR introduced in PETS 2023.
• We implement ChalametPIR introduced in CCS 2024, together with binary fuse filter.
• We implement PGM-index introduced in VLDB 2020.
• We implement various sorting algorithms in 2PC.
• We implement Plinko introduced in ePrint 2023. We introduce update API for client-preprocessing PIR.
• We implement Alternating-Moduli PRFs in CRYPTO 2024.
• We introduce offline/online paradigm for PSU.

Changed

• We refine implementations for Permutation Network with speedup.
• We refine implementations for Bloom Filter, Cuckoo Filter and Vacuum Filter with speedup (inspired by the Mobile Private Contact Discovery paper in USENIX Security 2019).
• We refine implementations for Piano and MIR PIR with speedup.
• We introduce UNSAFE for faster type convertions.
• We introduce iterator for BFV implementations in mpc4j-crypto-fhe.

Removed

• We remove mpc4j-common-jnagmp since we meet bugs when using JNA-GMP on Mac M3.
• We remove mpc4j-crypto-phe since PHE is no longer neceesary for mpc4j.
• We add functions supporting large data (>2G) transmissions in mpc4j-common-rpc.

Fixed

• We fix the bug in implementing SimplePIR. Now the rounding operation is implemented correctly.
• We fix the bug for generating $n \ge 2^{22}$ for PSO. Thanks Binbin Tu for reporting this bug.

v1.1.2

[1.1.2]

Added

• [important] Restrict that the minimal JDK version is JDK 21. We merge mpc4j-crypto-simd into corresponding modulues.
• Implement Half-tree introduced in EUROCRYPT 2023.
• Implement EACode and ECCode introduced in CRYPTO 2022 and CRYPTO 2023, respectively.
• Implement Secret-shared shuffle introduced in ASIACRYPT 2020.
• Implement altering module OPRF introduced in CRYPTO 2024.
• Speedup GF2E implementations.

Changed

• We fully re-organized PIR implementations and now they are more clear and easy to read.
• We fully re-organized PCG implementations, moving triple, daBit, OSN from mpc4j-s2pc-opf to mpc4j-s2pc-aby. Based on this modification, we implement a TrustedDealer to deal with all PCG generations.

Fixed

• Fixed a bug for the implementation of order-preserving encryption (EUROCRYPT 2009). Now that order-preserving encryption supports any output range $\mathcal{R}$ conditioned on $|\mathcal{R}| \geq |\mathcal{D}|$ where $|\mathcal{D}|$ is the input range.
• Fixed a bug for PSU implementations. Now we can correctly verify if PSU is implemented correctly.

Release 1.1.1

[1.1.1]

Added

• [important] Create a new module mpc4j-crypto-simd, introduce Vector API to implement SIMD operations for bit matrix transpose. However, this requires to develop mpc4j using JDK 17 or later.
• Create a new module mpc4j-crypto-algs, implement order-preserving encryption (EUROCRYPT 2009).
• Create a new module mpc4j-s3pc-abb3, start to implement honest-majority three-party protocols.
• Now mpc4j automatically compresses equal-length data packet for NettyRpc.
• Implement Waksman networks.
• Implement unbalanced private set union proposed in CCS 2023.

Changed

• update the package format so that now mpc4j allows many sub-protocols.
• update the implementation of BandLinearSolver. Now the implementation only needs linear memory cost.
• re-organize OKVS implementation, remove many unnecessary codes.

Fixed

• Fix a bug that FileRpc wrongly counts the communication cost.
• Fix a bug that reports invalid pointer when running examples. The bug comes from MCL. Considering the fact the OpenSSL also provides asm implementations for ECC, we now remove MCL.

Release 1.1.0

[1.1.0]

Added

• [important] mpc4j-crypto-fhe
  • Create a new module mpc4j-crypto-fhe to add implementations for FHE.
  • Implement BFV scheme. The implementation can be seen as a Pure-Java version of SEAL.

Changed

• [important] mpc4j-common-structure

  • rename mpc4j-crypto-matrix to mpc4j-common-structure and re-organize codes. Move Filter and LPN from mpc4j-common-tool to mpc4j-common-structure.
  • Update serialization methods for Filters to same communication costs.

• mpc4j-common-tool

  • Optimize reduceByteArray in BytesUtils.

• mpc4j-s2pc-pir

  • Refine code for SimplePIR.

Fixed

• mpc4j-s2pc-opf
  • Fix a bug that OprfUtils generates wrong sets for unequal set size.
• mpc4j-s2pc-pir
  • Fix a security flow in the implementation of Labeled PSI (CCS 2021).
• mpc4j-s2pc-pso
  • Fix some bugs and refine codes for many PSI implementations.

v1.0.9

Added

• mpc4j-common-tool
  • Introduce the way of setting ball-and-box argument in open source code VOLE-PSI, see MaxBinSizeUtils for more details.
  • Introduce a more efficient way of doing operations in GF128 field. The implementation is inspired by the blog Reversing a Finite Filed Multiplication Optimizaion.
  • Implement operations in GF64 field.
• mpc4j-common-matrix
  • Implement "Blazing Fast OKVS" introduced in the paper "Blazing Fast PSI from Improved OKVS and Subfield VOLE". The implementation is inspired by the open-soure code VOLE-PSI.
  • Implement "band encoding OKVS" introduced in the paper "Near-Optimal Oblivious Key-Value Stores for Efficient PSI, PSU and Volume-Hiding Multi-Maps". We thank Joon Young Seo and Kevin Yeo for the offline discussion of some implementation details.
• mpc4j-s2pc-pcg
  • Implement silent VOLE (both for semi-honest version and the malicious version) in GF128 field, using the technique introduced in the paper "Wolverine: Fast, Scalable, and Communication-Efficient Zero-Knowledge Proofs for Boolean and Arithmetic Circuits".
  • Implement single-point OT / single-point VOLE for ease of tests.
• mpc4j-s2pc-opf
  • Implement private set membership protocol introduced in the paper "Circuit-PSI with Linear Complexity via Relaxed Batch OPPRF".
  • Implement VOLE-OPRF introduced in the paper "VOLE-PSI: fast OPRF and circuit-PSI from vector-OLE".
• mpc4j-s2pc-pir
  • Implement native and PBC batch query for index PIR.
  • Implement unbalanced circuit PSI introduced in the paper "PSI with computation or Circuit-PSI for Unbalanced Sets from Homomorphic Encryption".
  • Implement circuit PSI (both for equal-size and unequal-size) introduced in the paper "Efficient circuit-based PSI with linear communication".
  • Implement circuit PSI (both for equal-size and unequal-size) introduced in the paper "Circuit-PSI with Linear Complexity via Relaxed Batch OPPRF".
  • Implement client-preprocessing PIR introduced in the paper "Piano : Extremely Simple , Single-Server PIR with Sublinear Server Computation". The implementation is inspired by the open-source code Piano-PIR.
  • Implement client-preprocessing PIR introduced in the paper "Simple and Practical Amortized Sublinear Private Information Retrieval".
• mpc4j-s2pc-pso
  • Implement aider-PSI introduced in the paper "Scaling private set intersection to billion-element sets".
  • Implement RT21 PSI introduced in the paper "Compact and Malicious Private Set Intersection for Small Sets". The implementation is inspired by the open-source code MiniPSI.
  • Implement PRTY19 PSI introduced in the paper "SpOT-Light : Lightweight Private Set Intersection from Sparse OT Extension".
  • Implement PRTY20 PSI introduced in the paper "PSI from PaXoS: Fast, Malicious Private Set Intersection".
  • Implement DCW13 PSI introduced in the paper "When private set intersection meets big data: An efficient and scalable protocol".
  • Implement RS21 PSI introduced in the paper "VOLE-PSI: fast OPRF and circuit-PSI from vector-OLE".
  • Implement RR22 PSI introduced in the paper "Blazing Fast PSI from Improved OKVS and Subfield VOLE".
  • Implement PSZ14 PSI introduced in the paper "Faster Private Set Intersection based on OT Extension".

Changed

• mpc4-common-tool
  • Introduce ways of computing distinct hashes in the open-source code VOLE-PSI (related to Bloom Filter, Garbled Bloom Filter, and Garbled Cuckoo Table).
  • Choose parameters for no-stash cuckoo hash for small item sizes.
• mpc4j-common-matrix
  • Refactor codes for OKVS so that OKVS implementations with doubly obliviousness share the same code with standard OKVS implementations.
• mpc4j-s2pc-pir
  • Faster matrix multiplication by avoiding unnecessary module operation in SimplePIR.
  • Refined labeled-PSI implementations based on the open-source code APSI of the paper "Labeled PSI from Homomorphic Encryption with Reduced Computation and Communication".

Fixed

• common
  • Update documentations to show to install FourQ, and how to solve the problem if FourQ test cases are failed.
• mpc4j-s2pc-pcg
  • Fix a bug for wrong LPN parameters used in silent OT.
  • Fix a bug for malicious-secure 1-out-of-2 COT introduced in the paper "SoftSpokenOT: Communication - Computation Tradeoffs in OT Extension" and fixed by the revised version of the paper "Actively Secure OT Extension with Optimal Overhead".
• mpc4j-s2pc-pir
  • Fix a bug for SimplePIR to support values with arbitrary bit length (instead of bit length that divides Byte.SIZE).
  • Fix a bug for Vectorized PIR to support values with arbitrary bit length (instead of bit length that divides Byte.SIZE).
• mpc4j-s2pc-pjc
  • fix a bug when running PID with unequal set size.

v1.0.8

Added

• mpc4j-common-circuit
  • We abstract MpcZlVector and MpcZlParty.
  • We add some circuit implementations, including adder, multiplier, and sorting network.
• mpc4j-common-tool
  • We add a new BitVector named CombinedBitVector that tries its best to support efficient BitVector operations. Now users can use CombinedBitVector for all cases.
• mpc4j-crypto-matrix
  • We implement Zp matrix and Zp64 matrix.
• mpc4j-s2pc-pcg
  • We implement coin-tossing protocols with semi-honest and malicious security.
• mpc4j-sp2c-aby
  • We implement Trust-Dealer model, that is, an aider can distribute Boolean and Multiplication triples for general MPC.
  • We implement daBits and EdaBits.
  • We implement some comparisons.
• mpc4j-s2pc-opf
  • We implement Naor-Ringold OPRF, and OPRP-based OPRF.
• mpc4j-s2pc-pir
  • We implement more index PIRs, including Simple / Double PIR (USENIX Security 2023), Mul PIR (USENIX Security 2021), constant-weight PIR (USENIX Security 2022).
  • We implement Pantheon PIR (VLDB 2022).
• mpc4j-s2pc-pso
  • We formalize and implement some PSI cardinality protocols, including EC-DH-based, DH-OPRF-based, and circuit-PSI-based.
  • We implement server-aided PSI protocols.
• others
  • We add test cases for PSU, PID and PMID.

Changed

• mpc4j-common-tool
  • We refine Filter implementations.
  • We refine implementations for sparse bit vector and sparse bit matrix.
• mpc4j-s2pc-pcg
  • We add silent model for all protocols that can leverage silent OT to reduce communication costs.
  • We remove number of bits / number of elements in general MPC.
• mpc4j-crypto-matrix
  • We move OKVS implementations into mpc4j-crypto-matrix.
• others
  • We refine configs for multi-party protocols so that we can remove many duplicate codes.
  • We refine test cases for multi-party protocols so tha we can remove many duplicate codes.

Fixed

• mpc4j-common-tool
  • We fixed a bug in CommitFactory. We need to create a commitment scheme with SHA256 hash for STANDARD and with SM3 hash for INLAND.
• mpc4j-crypto-matrix
  • We fixed a bug for toString() in vectors and databases. We need to correctly display the string even if the vector (the database) is empty (with num = 0).

v1.0.7

[1.0.7]

We are sorry that there is a missing file (OpBoostPtoConfig in mpc4j-sml-opboost) in previous v1.0.7.

Added

• mpc4j-common-circuit
  • We add a new module mpc4j-common-circuit to write all circuits in a unified manner.
  • We add some basic integer circuits: add, sub, increase one, equality (eq), less than or equal to (leq).
• mpc4j-crypto-matrix
  • We add a new module mpc4j-crypto-matrix to put functionalities related to cryptographic matrix operations.
  • We add some database / vector implementations.
• mpc4j-common-rpc
  • We add receiveAny() in Rpc.
  • We update the way of generating taskId. Now all sub-protocols have the same taskId with the root protocol. We distinguish sub-protocols using encodeTaskId. See AbstractMultiPartyPto for more details.
• mpc4j-common-tool
  • We add algebra operation interfaces in galoisfiled, including zl (Z mod (1 << l)), zl64 (Z mod (1 << l) where l < 64), zn (Z mod n), zn64 (Z mod n where n < (1 << 64)), zp (Z mod p where p is a prime), zp64 (Z mod p where p is a prime and p < (1 << 64)).
  • We introduce FourQ ECC.
• mpc4j-dp-service
  • Now main supports more configurations: (1) Allow running without plain case; (2) Allow no/empty settings for α, ε_w, fo_types, hg_types.
  • Add necessary test cases for HhLdpMain.
• mpc4j-s2pc-pcg
  • We add HE-based and OT-based multiplication triple generation protocols introduced in the DSZ15 paper.
  • We add FHE-based multiplication triple generation protocol introduced in the RSS19 paper.
  • We implement pre-computed 1-out-of-n OTs based on the silent OT.
• mpc4j-s2pc-aby
  • We refine many implementations for Boolean circuits.
  • We implement mux operations introduced in RRK+20 and RRG+21 papers.
  • We implement Boolean circuit based PEQT protocol and the optimized PEQT protocol introduced in the CGS22 paper.
• mpc4j-s2pc-pir
  • We implement vector PIR introduced in the MR23 paper.
• mpc4j-s2pc-opf
  • We create a new module mpc4j-s2pc-opf for oblivious pseudo-random functions.
  • We implement programmable OPRFs based on OKVS introduced in the PSTY19 paper.
  • We implement related-batch programmable OPRFs introduced in the CGS22 paper.
  • We implement single-query OPRF introduced in the RA17 paper.
• mpc4j-s2pc-pso
  • We implement two circuit PSI protocols (without associated payload) introduced in the PRTY19 and CGS22 paper.

Fixed

• mpc4j-common-tool
  • Fix a bug when switching the elliptic curve. In Missing docs for c++ interface? #72, the MCL author said "The current version does not support multi parameters. At first, I had developed the features, but I gave up it because a class dependency was very complicated." It brings some problems when we want to switch from an elliptic curve to another one that both use MCL. Now, we only allow users to use SEC_P256_K1 with MCL.
• mpc4j-dp-service
  • Fix a bug for AppleHcmsFoLdp, we note that in Java, a % b (for b > 0) can have negative value. Therefore, we need to write Math.abs(a % b) instead of directly a % b to ensure a % b must be in [0, b). Thank Xiaochen Li for the report.
  • Fix a bug for OLH and FLH, we note that $g$ in OLH and FLH must be an integer. Therefore, we cannot directly use the optimized frequency estimation formula to estimate the count. Instead, we use the original formula.
• mpc4j-s2pc-pcg
  • We slightly reduce the communication cost for distributed oblivious puncturable OPRF.
• mpc4j-s2pc-aby
  • Now we allow large BitNums per operations in the Boolean circuit.

v1.0.6

[1.0.6]

Added

• mpc4j-common-sampler
  • We implement many discrete Gaussian sampling techniques, including native sampling, Alias sampling, sigma-2 sampling, convolution techniques, and discrete gaussian sampling introduced in NIPS 2020.
• mpc4j-common-tool
  • We implement metrics used for HeavyHitter (in metrics/HeavyHitterMetrics.java), including NDCG (Normalized
    Discounted Cumulative Gain), precision, and relative error.
  • We introduce a new tool named BitVector for efficient bit operations.
  • We add MathPreconditions for math precondition checks.
  • We implement the non-cryptographic hash function BobHash and introduce xxHash in pure-Java.
• mpc4j-dp-service
  • We create a new module mpc4j-dp-service for implementing specific differential private mechanisms, e.g., Frequency Oracles.
  • We implement state-of-the-art LDP-based frequency oracle mechanisms, including Hadamard-related mechanisms, Unary Encoding (UE)-related mechanisms, Direct Encoding (DE)-related mechanisms, Local Hash (LH)-based mechanisms.
• mpc4j-s2pc-pir
  • We implement SealPIR, OnionPIR and FastPIR.
• mpc4j-s2pc-pjc
  • We create a new module mpc4j-s2pc-pjc to manage "Private Join and Compute" protocols, such as PSI-CA, PID, PMID,
    PSI-CA-SUM, and others.

Changed

• common
  • Previously, we place our own log4j.properties in resources. However, this may reject developers to use its
    own log4j.properties. We replace all log4j.properties from main/resources to test/resources.
  • We optimize LongUtils.ceilLog2 and some implementations in BigIntegerUtils based on Guava.
• mpc4j-common-tool
  • We rename package correlation to metrics so that we can include other metrics in that package.
  • We replace RankUtils.java with package util.
  • We optimize implementations for the Hadamard matrix and the Hadamard coder.
• mpc4j-s2pc-pso
  • We move blackIP data from module mpc4j-s2pc-pso to the dictionary data.
  • We move PID and PMID from module mpc4j-s2pc-pso to module mpc4j-s2pc-pjc.

Fixed

• mpc4j-common-tool
  • We fixed a bug in RandomCoderUtils.java, thanks Qixian Zhou for reporting.

v1.0.5

v1.0.5 released! We have the following updates:

• We add batched polynomial implementation algorithms (both for Java and C/C++) introduced in the CRYPTO 2019 paper SpOT-Light: Lightweight Private Set Intersection from Sparse OT Extension.
• We now support the pure-Java Ristretto curve. We also support pure-Java Elliagtor encoding/decoding introduced in the CCS 2021 paper Compact and Malicious Private Set Intersection for Small Sets.
• We implemented XPIR proposed in the PETS 2022 paper XPIR: Private Information Retrieval for Everyone.
• We add two PSU applications introduced in our PSU paper, namely IP blacklist aggregation and Private ID.