Bitcoin-NG: Scalable Blockchain Protocol

• Bitcoin-NG is a blockchain protocol that overcomes scalability limits by decoupling leader election (via PoW key blocks) from rapid transaction serialization through microblocks.
• It employs a two-layer design where key blocks elect a leader and microblocks enable high-frequency transaction processing, optimizing both throughput and latency.
• The protocol incorporates nuanced fee-splitting and security analyses to ensure incentive compatibility and mitigate attacks such as selfish and greedy mining.

Bitcoin-NG is a blockchain protocol designed to overcome the inherent scalability limits of Bitcoin’s Nakamoto consensus, achieving near-optimal throughput and network-limited latency by decoupling leader election from transaction serialization. The protocol features a two-plane design: key blocks, mined via proof-of-work (PoW), periodically select leaders who subsequently serialize transactions into microblocks at rates constrained only by network capacity. Bitcoin-NG has inspired several high-performance blockchain designs and introduces nuanced incentive structures and rigorous analyses for protocol security and incentive compatibility (Eyal et al., 2015, Niu et al., 2020, Hu et al., 2023).

1. Architectural Principles: Key Blocks and Microblocks

Bitcoin-NG divides time into sequential epochs, each initiated by the mining of a key block. Key blocks play no direct role in transaction serialization but instead serve as leader-election artifacts, with PoW acting as an unpredictability source. Upon mining a key block, the miner becomes the epoch leader and gains exclusive authority to issue a stream of cryptographically signed microblocks, which serialize transactions at high frequency (e.g., one every 20 seconds) without further PoW (Eyal et al., 2015, Niu et al., 2020).

Key elements of the protocol structure are:

• Key Blocks: Contain pointers to the parent key block, metadata Merkle root, PoW nonce, and leader identifier.
• Microblocks: Carry batches of transactions, reference the previous microblock, and are signed by the epoch leader; they confer no chain weight and do not influence future leader selection.
• Epochs: Span the interval from one key block to the next, during which the leader serializes transactions via microblocks.
• Chain Selection: The main chain is defined by key blocks with the greatest cumulative PoW; microblocks on pruned key-block branches are orphaned.

This separation enables Bitcoin-NG to scale throughput up to node and network bandwidth limits, while preserving confirmation latency close to the network propagation delay plus the key-block interval (Eyal et al., 2015).

2. Incentive Mechanisms and Fee Splitting

Bitcoin-NG’s incentive mechanism ensures both robust security and optimal protocol operation by employing a fee-splitting approach:

• Key-Block Rewards: Each key block miner receives the entire block subsidy upon block confirmation in the main chain.
• Microblock-Transaction Fees: Transaction fees collected in microblocks during an epoch are split between the current and next leader, with fractions $r$ and $1-r$, respectively, where $0Niu et al., 2020). This model aligns the leader’s incentive for both prompt microblock publication and orderly handoff to the succeeding leader.

For regular transaction throughput, optimal security is obtained by setting $r$ within the bounds $\alpha < r < \beta$, where $\alpha$ is the attacker’s mining power and $\beta=1-\alpha$. For rare, high-fee transactions ("whale" transactions), more nuanced bounds apply, accounting for unsaturated microblock capacity.

Parameter selection must carefully balance attack resistance and efficient operation. For $\alpha<25\%$, $r=40\%$ is shown to be sufficient to deter attacks involving microblock withholding or rejection (Eyal et al., 2015, Niu et al., 2020).

3. Adversarial Models and Attack Analysis

Bitcoin-NG’s incentive analysis examines both previously known and novel adversarial strategies:

• Selfish Mining: Analysis via Markov decision processes (MDPs) reveals that the threshold for profitable selfish mining remains $\approx23.21\%$—identical to Bitcoin—provided $r$ is optimally chosen. Only when $\alpha \gtrsim 35\%$ does Bitcoin-NG yield marginally higher relative revenue to a selfish miner due to fee-splitting, but the revenue gain is minimal (Niu et al., 2020).
• Microblock Attacks: Two classes are considered:
  • Transaction-Inclusion Attack: Selfish miners withhold a fraction $\rho$ of their microblocks. The attacker's long-term revenue is governed by $$u_{\text{TI}} = \frac{\alpha – r \alpha \beta \rho}{1 – \alpha \beta \rho}$$.
  • Longest-Chain-Extension Attack: Selfish miners ignore honest microblocks, yielding $$u_{\text{LC}} = \frac{\alpha – (1−r) \alpha \beta \rho}{1 – \alpha \beta \rho}$$.
  • Incentive compatibility demands that $u_{\text{TI}}, u_{\text{LC}} \leq \alpha$, leading to $\alpha < r < \beta$ (Niu et al., 2020).
• Greedy-Mine Attack: Introduced in recent work, this attack targets high-fee (whale) transactions using a refined MDP model. The adversary can achieve greater expected reward than honest mining if $\alpha$ exceeds a threshold $\alpha_{\min}(\gamma)$, which is influenced by the propagation factor $\gamma$. For $\gamma=1$, $\alpha_{\min}\approx18\%$; for $\gamma=0.5$, $\alpha_{\min}\approx25\%$; for $\gamma=0$, no profitable attack exists until $\alpha>36\%$ (Hu et al., 2023).

Attack Type	Revenue Advantage Threshold	Revenue Expression
Selfish Mining (general)	$\alpha > 23.21\%$ (with optimal $r$)	Slight, for $\alpha \gtrsim 35\%$
Transaction-Inclusion Attack	$\alpha < r < \beta$ needed for no advantage	$u_{\text{TI}}$ as above
Greedy-Mine (whale-tx targeting)	$\alpha > 0.18$ if $\gamma=1$	$V_{\text{G}}(\alpha,\gamma) > V_{\text{H}}(\alpha)$

Deviation from the prescribed fee-splitting bounds can expose the protocol to profitable attacks, especially when adversarial mining power straddles these thresholds (Niu et al., 2020, Hu et al., 2023).

4. Network Capacity and Performance Scalability

Bitcoin-NG's transaction serialization via microblocks decouples throughput from consensus, allowing the effective transaction rate to scale linearly with the network’s microblock capacity. Letting $v$ denote the microblock issue rate and $f$ the key-block PoW rate, peak throughput approaches $\mathcal{O}(v)$, constrained by network bandwidth and node processing (Eyal et al., 2015, Niu et al., 2020).

• Network Model: Key blocks arrive as a Poisson process ($f$), yielding exponentially distributed inter-key-block times ($Y_i \sim \text{Exp}(f)$); microblocks are regularly emitted during each epoch at maximum sustainable rate $v$.
• Latency: Near-optimal confirmation latency is achieved: approximately a key-block interval plus one microblock propagation time.
• Attack Surface: The effectiveness of attacks does not increase as $v$ decreases due to bandwidth constraints; both honest and adversarial fee rates scale down equally, preserving the ratio of rewards and unchanged incentive bounds.

Empirical results show Bitcoin-NG achieves consensus delay ($\approx$1 s), time-to-prune, fairness, and mining-power utilization near their theoretical maxima under realistic network conditions. For example, with 1 MB key blocks and microblock intervals of 0.1–1 s, fairness and utilization both approach 1 (Eyal et al., 2015).

5. Advanced Incentive Analysis: MDP Joint Modeling

To capture the full complexity of miner incentives, recent work models the system as an MDP $(S,A,P,R)$ over the state of both key and microblocks:

• State Space: Specified as $(l_a, l_h, \text{fork}, \mu)$, where $l_a$ is attacker’s private chain lead, $l_h$ the public chain's lag, fork state, and microblock inclusion/exclusion status.
• Actions: Wait, adopt, override, match, revert, as dictated by the evolution of key-block races and microblock choices.
• Transitions: Determined by PoW competition ($\alpha$ for attacker, $1-\alpha$ for honest miners), with tie-breaking governed by parameter $\gamma$ (fraction of honest hashpower adopting the attacker's block in a fork).
• Rewards: Key-block miners receive $R_b$ per confirmed key block, microblock-epoch leaders receive transaction fees at rate $v/f$ split by $r$.

Numerical value-iteration over this model reproduces classical selfish-mining thresholds and identifies marginally increased revenue for attackers above $\alpha\gtrsim35\%$, but with little practical effect at lower adversary shares (Niu et al., 2020, Hu et al., 2023).

6. Security Bounds and Protocol Adjustments

While Bitcoin-NG, with optimally set $r$, matches Bitcoin’s selfish-mining security threshold, protocol vulnerabilities arise if mining power or network conditions deviate:

• Whale Transaction Vulnerability: The introduction of the Greedy-Mine attack demonstrates that for certain parameter regimes, adversaries can exploit the fee-splitting rule to profitably deviate for rare, high-fee transactions, lowering the minimum mining power required for profitable attack relative to standard selfish mining (Hu et al., 2023).
• Countermeasures: Raising the fee hand-off fraction ($r$) increases resistance: setting $r > 0.40$ ensures incentive compatibility against Greedy-Mine up to the corresponding mining power thresholds. Full elimination of next-leader fees ($r=1$) parallels the classical Bitcoin model but alters the distribution of transaction fees and may affect other protocol incentives.
• Protocol Flexibility: Honest mining remains strictly dominant if $$V_{\rm H}(\alpha;r^{\rm new}) \geq \max_{\text{attacks}} V_{\text{Attack}}(\alpha,\gamma;r^{\rm new})$$ for all $\alpha < 0.5$. Progressive deployment (e.g., at hard fork) and strict earliest-tip preference mitigate residual vulnerabilities (Hu et al., 2023).

7. Comparative Summary and Open Issues

Property	Bitcoin	Bitcoin-NG (with optimal $r$)
Security threshold	$\approx23.21\%$	$\approx23.21\%$ (can be lower if $r$ mis-set)
Throughput	PoW block-limited	$\mathcal{O}(v)$, network-limited
Confirmation latency	Block interval	Key-block interval + microblock delay
Fairness, mining utilization	Lower at higher block rates	Near-ideal at high microblock rates
Whale-tx attack resistance	Robust	Requires elevated $r$

Bitcoin-NG realizes near-optimal blockchain scalability by architectural decoupling and specialized incentive engineering but demands careful fee parameterization and fork-resolution logic to preserve security guarantees in the presence of diverse and adaptive adversarial mining strategies (Eyal et al., 2015, Niu et al., 2020, Hu et al., 2023). The protocol’s ongoing evolution centers on refining the reward structure (potentially dynamically) and improving fork resolution, opening directions for robust large-scale deployment.