[Paper Review] Consensus Mechanism Design based on Structured Directed Acyclic Graphs

Jiahao He, Guangju Wang+2 more|arXiv (Cornell University)|Jan 9, 2019

Blockchain Technology Applications and Security8 references4 citations

TL;DR

This paper proposes a consensus mechanism using a structured directed acyclic graph (DAG) to scale blockchain capacity without sacrificing security or decentralization. By embedding a Nakamoto chain within a strongly connected DAG and using a transaction assignment method that minimizes duplicate processing, the system achieves up to 1,179.6 transactions per second with a total latency of approximately 1,021 seconds (17 minutes), including 810 seconds of secure latency under 30% malicious hashing power.

ABSTRACT

We introduce a structure for the directed acyclic graph (DAG) and a mechanism design based on that structure so that peers can reach consensus at large scale based on proof of work (PoW). We also design a mempool transaction assignment method based on the DAG structure to render negligible the probability that a transaction being processed by more than one miners. The result is a significant scale-up of the capacity without sacrificing security and decentralization.

Motivation & Objective

• To address the scalability bottleneck in blockchain systems, which limits transactions per second (TPS) despite high demand.
• To maintain strong security and decentralization while increasing throughput beyond chain-structured blockchains.
• To reduce transaction confirmation latency and improve fee inclusion fairness for low-fee transactions.
• To deconcentrate mining power by enabling continuous block creation and reducing reliance on mining pools.
• To ensure that transactions are processed with negligible probability of being handled by multiple miners through a DAG-based mempool assignment method.

Proposed method

• Designing a structured, strongly connected DAG that embeds a Nakamoto chain to preserve security guarantees from proof-of-work (PoW).
• Using a mempool transaction assignment method based on DAG structure to minimize the chance of a transaction being processed by more than one miner.
• Modeling milestone creation as a Poisson process with a stochastic tag system (0 or 1) to represent honest or malicious block creation.
• Applying a probabilistic model to estimate the long-run proportion of milestones tagged with 1, based on Poisson arrival rates and transmission delay.
• Simulating failure frequency over 1 million sample paths to determine required secure latency for desired security levels (e.g., <10^-3 failure rate).
• Decomposing total latency into three components: queueing latency (W₁), infection latency (W₂), and secure latency (W₃), each modeled using analytical and simulation-based approaches.

Experimental results

Research questions

• RQ1 How can a DAG-based structure be designed to maintain the security of Nakamoto consensus while enabling parallel transaction processing?
• RQ2 What is the maximum achievable throughput of a DAG-based consensus system under PoW, and how does it scale with network parameters?
• RQ3 How can transaction assignment be optimized in a DAG to minimize duplicate processing and improve efficiency?
• RQ4 What is the required secure latency to ensure a failure probability below 10^-3 under varying levels of adversarial hashing power?
• RQ5 How do queueing, infection, and secure latencies combine to determine the total transaction confirmation time in the proposed system?

Key findings

• The system achieves a maximum theoretical throughput of approximately 1,179.6 transactions per second, based on 1,000 honest miners each producing 1.2 blocks per second and a mempool assignment efficiency of 98.3%.
• With 30% of hashing power controlled by malicious miners, a secure latency of 810 seconds is required to ensure a failure probability below 10^-3.
• The total transaction confirmation latency is approximately 1,021 seconds (just over 17 minutes), composed of 188 seconds of queueing latency, 23 seconds of infection latency, and 810 seconds of secure latency.
• The long-run average proportion of milestones tagged with 1 (indicating honest block creation) is estimated at 92.8% when 10% of miners are malicious, under the given parameters.
• Simulation results show that the failure frequency decays rapidly with increasing secure latency, with 130 seconds required for 10% adversarial power and 810 seconds for 30% adversarial power to achieve a failure rate below 10^-3.
• The proposed DAG structure enables continuous block creation with small block sizes (one transaction per block), reducing cryptographic puzzle difficulty and improving network responsiveness.