This recognition further reinforces the mission of our Sydney Blockchain Centre (lnkd.in/g2M7Y7iG), to advance innovation in blockchain and decentralised science.

This could have a variety of applications in Central Bank Digital Currency (our work was motivated by the architecture proposed by the Bank for International Settlements and the Hong Kong Monetary Authority), trusted hardware–based wallets, and beyond.

We propose Mosaic: client-side, local shard decisions, 20,000× faster, 228B input vs GBs, 95% throughput retained.

SOTA schemes like TxAllo (ICDE’23) use miner-driven global optimisation, requiring miners to sync the full ledger — aiming to cut costly cross-shard txs, but undermining sharded state’s goal of scalable performance.

We close both gaps:
• Constant latency (3Δ) in the sleepy model
• Graded Common Prefix (GCP): generic DAG finality in 2 steps (vs 4 in BFT)
• adding flexibility via extended Ebb-and-Flow

SOTA DAG protocols fall into two camps:
• Prioritise liveness: support dynamic availability, but with non-constant latency
• Prioritise safety: rely on protocol-specific BFT for finality

Achilles:
• Removes rollback prevention from the critical path
• Matches CFT protocols in latency (4 steps) & linear message complexity
• Delivers state-of-the-art performance among TEE-assisted BFT protocols

SOTA TEE-assisted BFT faces a trade-off:
• High performance (e.g. FlexiBFT) reduces fault tolerance
• Strong tolerance (e.g. Damysus) may suffer from rollback issues (e.g. if using SGX)