Non-custodial on-chain trading platforms are foundational to decentralized finance (DeFi), enabling transparent and verifiable asset exchanges while allowing users to retain direct control over their funds. However, scalability limitations on Layer 1 (L1) blockchains have driven the adoption of Layer 2 (L2) solutions, with zero-knowledge rollups (zkRollups) emerging as a prominent approach. Although zkRollups address scalability challenges, they encounter significant computational bottlenecks during proof generation, leading to elevated hardware requirements and extended proving times. This thesis presents VEX, an application-specific zkRollup architecture for verifiable exchange systems. To mitigate prover inefficiencies, two key optimizations are introduced: (1) Adopting PLONK with Se...[ Read more ]

Non-custodial on-chain trading platforms are foundational to decentralized finance (DeFi), enabling transparent and verifiable asset exchanges while allowing users to retain direct control over their funds. However, scalability limitations on Layer 1 (L1) blockchains have driven the adoption of Layer 2 (L2) solutions, with zero-knowledge rollups (zkRollups) emerging as a prominent approach. Although zkRollups address scalability challenges, they encounter significant computational bottlenecks during proof generation, leading to elevated hardware requirements and extended proving times. This thesis presents VEX, an application-specific zkRollup architecture for verifiable exchange systems. To mitigate prover inefficiencies, two key optimizations are introduced: (1) Adopting PLONK with Segment Lookup for Branching Logic, which dynamically activates transaction-specific sub-circuits based on batch inputs, ensuring that prover costs scale quasi-linearly with the number of active sub-circuits; and (2) Shared Logic Separation, which precomputes and independently proves shared operations, such as signature verification and hashing, reducing segment lookup overhead and enabling tailored optimizations. The proof components for shared and transaction-specific logic are aggregated through an AND composition of CP-SNARKs. A Link Protocol with universal setup is proposed, ensuring constant proof size and verification time for CP-PLONK compositions. Additionally, cross-framework composition between PLONK with Segment Lookup and Groth16 is facilitated through an intermediate PLONK circuit. Experimental evaluations demonstrate that VEX achieves a throughput of 11 transactions per second (TPS) on a budget machine with four physical cores and a batch size of 1024. Transaction-specific logic proving is 43 times faster than that of monolithic designs. These results highlight the potential of VEX to deliver significantly higher throughput with improved hardware configurations, positioning it as a scalable solution for verifiable exchange systems.