TX-SHIELD – Institutional Privacy Infrastructure

TX-SHIELD develops regulated privacy-preserving infrastructure for institutional payments (TX-SHIELD), private AI collaboration (OpenTMP LLM), and MPC-TSS key management (Collab-Key). TX-SHIELD’s three core solutions — TX-SHIELD, OpenTMP LLM, and Collab-Key — address compliance, compute, and custody privacy challenges faced by regulated institutions.

What it is

TX-SHIELD builds modular privacy layers/solutions for institutional finance and AI systems:

TX-SHIELD — a regulated private payment layer/solution built on MPC-based encryption and threshold key control.
It enables confidential settlement for stablecoins, RWAs, and bonds while ensuring regulator-auditable transparency.
Transactions are visible only to stakeholders, while regulators can access details through granted audit keys.
OpenTMP LLM — is a distributed edge AI training and inference framework designed for privacy-preserving large-language model collaboration. It combines federated learning and multi-party computation (MPC-FL) to keep data local while enabling encrypted aggregation and joint model updates. It powers collaborative, effcient, secure, and governable AI training across distributed environments.
Collab-Key — is a high-performance MPC-TSS framework integrated with CrossBar’s EMPC + ReRAM hardware to provide a unified, privacy-preserving stack for institutional custody and payments. By executing multi-party ECDSA signing within a ReRAM-backed secure element, the solution ensures that full private keys are never reconstructed, eliminating single points of failure at both algorithmic and physical layers. Featuring an offline-by-default architecture and FIDO2 authentication, this integration provides high-assurance protection and regulator-auditable transparency for stablecoins, RWAs, and digital bonds.

Each module can operate independently or as part of a unified privacy-preserving stack across payments, compute, and custody.

Fits with patterns (names only)

TX-SHIELD:

pattern-private-stablecoin-shielded-payments
pattern-private-pvp-stablecoins-erc5753
pattern-regulatory-disclosure-keys-proofs
pattern-te-ex-settlement

OpenTMP LLM (adjacency):

pattern-adj-private-consortium-ledger
pattern-adj-delayed-post-trade-reporting

Collab-Key:

pattern-key-management-threshold-kms

Not a substitute for

TX-SHIELD:

ZK-based L2 privacy frameworks (e.g., Aztec, Scroll)
General-purpose MPC or TEE frameworks for secure computation (TX-SHIELD focuses on transactional privacy and compliance)
Traditional on-chain settlement systems without regulator access

OpenTMP LLM:

Centralized AI model training pipelines
Non-encrypted data-sharing frameworks

Collab-Key:

Single-key custodial wallets
Hardware-based key storage only

Architecture

TX-SHIELD

Implements a high performance MPC-based private payment layer with threshold key control and an audit-key protocol for regulator visibility.
Only sender, receiver i.e. stakers, and authorized regulators can access encrypted transaction details.
Optimized MPC execution enables high performance (~10k TPS).

OpenTMP LLM

Distributed AI architecture using federated learning and multi-party computation (MPC-FL) with threshold-secure aggregation.
Supports edge acceleration, model distillation, quantization, and joint model governance.

Collab-Key

Supports 2PC and multi-party ECDSA protocols, integrated with CrossBar EMPC + ReRAM hardware. Key shards are processed and stored within a ReRAM-backed secure element, ensuring private keys are never reconstructed in full at either the algorithmic or physical levels. Features an offline-by-default architecture with native FIDO2 support, integrating seamlessly with institutional KMS and APIs.

Privacy domains (if applicable)

Private Payments / Compliance Infrastructure
Collaborative AI / Federated Learning Privacy
ReRAM-backed Institutional Custody / Key Management

Enterprise demand and use cases

TX-SHIELD:

Institutional settlement for stablecoins, tokenized RWAs, and bonds on-chain.
Ideal for financial institutions needing confidentiality and compliance together.

OpenTMP LLM:

Privacy-preserving AI model training and inferences for enterprises and regulated sectors, such as healthcare, finance, and government.

Collab-Key: Institutional wallets, custodians, and enterprise-grade signing infrastructure requiring fault-tolerant, hardware-secure (ReRAM) key management to eliminate single points of failure.

Technical details

TX-SHIELD:

MPC-based encryption, threshold key control (TSS), high-throughput multi-party computation.

OpenTMP LLM:

MPC-FL, Distributed Learning, Edge AI Acceleration, SFT, RLHF.

Collab-Key:

MPC-TSS (ECDSA 2PC / multi-party), Threshold Signatures, Secure Key Generation.

Strengths

TX-SHIELD delivers institutional-grade transactional privacy with built-in regulatory visibility.
OpenTMP LLM enables privacy-preserving multi-party AI collaboration across edge and federated (hybrid on/off-chain) environments.
Collab-Key provides fault-tolerant, MPC-TSS-based signing with formal security foundations (USENIX Security 2025), strengthened by CrossBar ReRAM hardware-level isolation to eliminate physical-layer vulnerabilities.

Risks and open questions

TX-SHIELD:

Governance over regulator audit keys
Integration complexity across different blockchain environments

OpenTMP LLM:

Coordination complexity in multi-party settings
Trade-offs between model performance and full encryption overhead

Collab-Key:

Performance scaling with increased party count
Implementation complexity across heterogeneous custody systems
Challenges in standardizing the interface between the ReRAM-backed secure element and diverse institutional legacy KMS/custody environments.
Establishing protocols for secure shard migration or recovery in the event of physical hardware failure or decommissioning of the CrossBar EMPC module.

Links

Official Website: TX-SHIELD | BenPay|Privacy Wallet

Contact: ZYX Research | haiyangxue@smu.edu.sg

Papers: