Decentralised Dark Pool Matching on India's Open Network for Digital Commerce via Secure Multi Party Computation

India’s Open Network for Digital Commerce (ONDC), built on the Beckn open protocol, enables decentralised B2B wholesale trade for micro, small, and medium enterprises (MSMEs). However, its transparent broadcast architecture exposes buyer bids and seller asks to competing participants and the gateway, enabling order-book front-running that destroys the price advantage bulk trade is intended to deliver. We present the first privacy-preserving order-matching protocol for ONDC - a decentralised dark pool in which buyer bids and seller asks are never revealed in plaintext to any party, including the ONDC gateway. Our protocol composes Yao’s Garbled Circuits with 1-of-2 Oblivious Transfer (OT) over X25519 and is deployed across three non-colluding SMPC nodes consistent with the Beckn participant model. We extend the classical binary matching predicate b ≥ a to a fused heterogeneous lot-compatibility predicate covering non-uniform commodity quantities and partialfill thresholds - the first such extension in the MPC dark pool literature. We implement a complete Python prototype spanning all five protocol roles, garble and evaluate a 353-gate Boolean circuit at k = 32 bits, and demonstrate ≈30 ms end-to-end perpair latency with 96.2 KB per-round communication overhead. Protocol match rates are proven identical to plaintext oracles across all tested market scenarios. Our system constitutes a structural blueprint for privacy-preserving commerce on India’s digital public infrastructure.