PaperlayerView exact revision on arXiv

Source-linked AI summary

Taxonomy of Centralization in Public Blockchain Systems: A Systematic Literature Review

Ashish Rajendra Sai, Jim Buckley, Brian Fitzgerald, Andrew Le Gear

arXiv:2009.12542v1cs.CYcs.CR

TL;DR

Public-blockchain decentralization is not guaranteed by design, although it is central to the security rationale for these systems. This paper reviews the literature and refines a layered taxonomy with expert interviews, providing a framework for discussing centralization and its measurement.

  • Problem

    Public-blockchain decentralization is not guaranteed by design, making it important to understand centralization concerns relevant to security.

    Source: §1 · p. 2

  • Method

    The authors synthesize centralization research into a taxonomy organized by blockchain architecture and refine it through expert interviews.

    Source: §2.2 · p. 5, §6 · p. 25

  • Results

    The review produces an initial taxonomy, surveys measurement approaches, and identifies studies that lack centralization measurement techniques.

    Source: §6 · p. 25

  • Takeaways & Limitations

    The taxonomy provides a framework for more systematic discussion of centralization in major blockchain systems.

    Source: §6 · p. 25

    The literature search may have missed relevant work with different phrasing or work absent from the leading repositories, including grey literature.

    Source: §7.2 · p. 28

Abstract

from arXiv · show

Bitcoin introduced delegation of control over a monetary system from a select few to all who participate in that system. This delegation is known as the decentralization of controlling power and is a powerful security mechanism for the ecosystem. After the introduction of Bitcoin, the field of cryptocurrency has seen widespread attention from industry and academia, so much so that the original novel contribution of Bitcoin i.e. decentralization, may be overlooked, due to decentralizations assumed fundamental existence for the functioning of such cryptoassets. However recent studies have observed a trend of increased centralization in cryptocurrencies such as Bitcoin and Ethereum. As this increased centralization has an impact the security of the blockchain, it is crucial that it is measured, towards adequate control. This research derives an initial taxonomy of centralization present in decentralized blockchains through rigorous synthesis using a systematic literature review. This is followed by iterative refinement through expert interviews. We systematically analyzed 89 research papers published between 2009 and 2019. Our study contributes to the existing body of knowledge by highlighting the multiple definitions and measurements of centralization in the literature. We identify different aspects of centralization and propose an encompassing taxonomy of centralization concerns. This taxonomy is based on empirically observable and measurable characteristics. It consists of 13 aspects of centralization classified over six architectural layers Governance Network Consensus Incentive Operational and Application. We also discuss how the implications of centralization can vary depending on the aspects studied. We believe that this review and taxonomy provides a comprehensive overview of centralization in decentralized blockchains involving various conceptualizations and measures.

1 Introduction

Decentralization is not guaranteed by public-blockchain design, yet it underpins security and utility. The paper uses a layered review and expert refinement to organize centralization concerns and their measurement.

1 Introduction

  • Centralization trends in public blockchains raise security concerns because security guarantees depend on the honest-majority assumption.Centralization may also threaten adoption and uncensored transactions.Source: §1 · p. 2
  • Consensus-power concentration is commonly measured, but it does not capture governance, wealth concentration, or participant geography.The paper argues for a vocabulary that supports more holistic measurement across blockchain building blocks.Source: §1 · p. 2
  • The study asks which aspects of public-blockchain centralization exist and how centralization can be adequately measured.Source: §1 · p. 3
  • The authors review 89 articles from a five-step systematic literature review and refine the resulting taxonomy through ten expert interviews.The interviews included four academic researchers and six industry experts.Source: §1 · p. 3
  • The paper maps centralization aspects and measurement techniques into a taxonomy, applies it to Bitcoin and Ethereum, and identifies research gaps.The authors also report a lack of objective metrics for some centralization-causing factors.Source: §1 · p. 3

2 Background

Blockchains combine a linked data structure with distributed consensus and incentives. Consensus mechanisms shape participation and rewards, while public, private, and consortium systems differ in who may participate.

2 Background

  • A blockchain stores data in chronologically linked blocks secured by computationally hard hash puzzles, making alteration of past data increasingly expensive.The passage describes this growth in difficulty as providing deterministic immutability.Source: §2 · p. 3
  • Participants use a peer-to-peer network and incentives to agree on one state; prominent consensus mechanisms include Proof-of-Work and Proof-of-Stake.Source: §2 · p. 3
  • Proof-of-Work rewards the first participant to solve and propagate a puzzle, whereas Proof-of-Stake bases reward distribution on participants’ monetary assets.The passage notes that Proof-of-Work is criticized for electricity use.Source: §2 · p. 4
  • Public blockchains allow open consensus participation, while private and consortium systems restrict participation to users approved by a trusted authority.Source: §2 · p. 4

2.1 Decentralization and Public Blockchain

Public-blockchain decentralization distributes control among participants and supports security under an honest-majority assumption. The paper responds to fragmented, technically narrow accounts by proposing a taxonomy organized across blockchain layers.

2.1 Decentralization and Public Blockchain

  • Public-blockchain consensus relies on decentralized control to remain secure while a majority of network participants are honest.Source: §2.1 · p. 4
  • Prior reviews identify decentralization as a reason for blockchain adoption and document applications in finance, IoT, healthcare, and smart cities.Source: §2.1 · p. 4
  • Centralized consensus power is linked to threats including the 51% attack and selfish mining, which can harm the network at 26% consensus power.Source: §2.1 · p. 4
  • Because existing centralization research is fragmented and technical-only views can mislead, the paper develops a layered taxonomy refined through expert interviews.Source: §2.1 · p. 4

2.2 Architecture of Public Blockchains

The paper adopts a generic layered architecture to examine centralization holistically, from data storage and network communication through consensus and incentives to applications.

2.2 Architecture of Public Blockchains

  • A generic layered architecture is chosen to capture centralization factors beyond any one implementation, drawing on distributed-computing models.Earlier architecture descriptions often focused microscopically on selected components.Source: §2.2 · pp. 4–5

2.2.5 Contract Layer

  • The contract layer executes computational scripts for consensus and incentives; Bitcoin’s limited Script lacks Turing completeness, unlike Ethereum’s engine.Source: §2.2 · p. 5, §2.2.5 · p. 6

2.2.6 Application Layer

  • The application layer supports applications such as cryptocurrencies and DApps, while public APIs connect blockchain services to wallets and exchanges.Source: §2.2 · p. 5, §2.2.6 · p. 6

2.2.1 Data Layer

  • The data layer specifies the data structure and transaction storage, including cryptographic primitives required by participants and clients.Source: §2.2.1 · p. 5

2.2.2 Network Layer

  • The network layer governs node connections, intercommunication, peer discovery, and information dissemination.Source: §2.2.2 · p. 5

2.2.3 Consensus Layer

  • The consensus layer enables the network to agree deterministically on a single state of the ledger with a defined degree of assurance.Source: §2.2.3 · p. 5

2.2.4 Incentive Layer

  • The incentive layer distributes rewards, often block rewards, to support an honest majority and connects technical operations with user-facing layers.Source: §2.2.4 · pp. 5–6

3 Methodology

The study follows a four-phase systematic literature review to identify and measure centralization factors, map them to a blockchain architecture, and refine the resulting taxonomy through expert interviews. Its review process narrows a broad search to 89 articles for analysis.

3 Methodology

  • The methodology follows Kitchenham et al.’s systematic-review guidelines through search, selection, quality assessment, data extraction, and analysis.Source: §3 · p. 6, §3.1 · p. 7

3.1.1 Phase 1: Research Questions and Query Formation

  • The review asks which centralization aspects occur in public blockchains and which techniques measure them.For each factor, the researchers retained references to articles and recorded any measurement techniques used.Source: §3.1.1 · pp. 7–8
  • Researchers formulated initial search keywords from blockchain and centralization synonyms and alternate words, then snowballed two most-cited relevant studies [14] to select the top five additional keywords.Source: §3.1.1 · p. 8

3.1.2 Phase 2: Article Search and Selection

  • The search and selection process reduced 3,572 deduplicated articles to 212 relevant articles and then 89 articles for data extraction.Selection was restricted to English-language articles published after Bitcoin’s 2009 introduction; reviewer agreement on article inclusion and rejection had Cohen’s Kappa exceeding 0.8.Source: §3.1.2 · pp. 8–9

3.1.3 Phase 3: Data Extraction

  • The review recorded centralization factors and their associated measurement techniques from shortlisted articles as the building blocks of the taxonomy.Source: §3.1.3 · p. 9

3.1.4 Phase 4: Development of Initial Taxonomy

  • The review maps extracted centralization factors to blockchain architecture layers, refining the architecture when a factor cannot reasonably fit existing layers.The mapping identifies 63 articles discussing consensus-layer centralization, the highest count for any layer.Source: §3.1.4 · p. 9

3.2 Interview with experts

  • Ten expert interviews with four academic and six industry experts were used to validate and refine the initial taxonomy and architecture.The interviews used open-ended questions about centralization, factor identification and measurement, and taxonomy accuracy.Source: §3.2 · pp. 9–10

4 Taxonomy of Centralization of Public Blockchain

The taxonomy organizes public-blockchain centralization factors across architectural layers, adding governance and operational layers where the generic architecture did not capture relevant concerns. It defines centralization in terms of constraints on participants and reports 13 aspects across six layers.

4 Taxonomy of Centralization of Public Blockchain

  • Governance and operational layers were added because the generic architecture did not capture governance mechanisms or constraints on operating a node.The generic architecture’s Data and Contract layers were removed because surveyed articles identified no centralization aspects in them.Source: §4 · p. 10
  • Most taxonomy factors were well established in literature or interviews, while Bandwidth Concentration and Routing Centralization were possible exceptions.Although only one academic article mentioned Node Discovery Protocol Control, most interviewees considered it relevant.Source: §4 · p. 11
  • The taxonomy defines centralization as architectural aspects restricting most participants through direct or indirect economic, social, or technical constraints.Source: §4 · p. 11
  • The taxonomy reports 13 centralization aspects across six layers, with about two factors identified per resulting layer on average.Some factors, such as Wallet Concentration, had no proposed measurement technique.Source: §4 · pp. 10–11

4.1 Governance

The review and interviews identify governance as a centralization concern, dividing it into owner control and improvement protocol. They describe how concentrated ownership or control over proposed changes can affect consensus power, economic centralization, and network evolution.

4.1 Governance

  • 80% of interview participants identified governance as a significant centralization threat.Participants described decentralized governance as a vital component of public blockchains.Source: §4.1 · p. 11
  • The review and interviews divide governance centralization into owner control and improvement protocol.These categories are presented in Table 4.Source: §4.1 · p. 12, Table 4 · p. 13

4.1.1 Owner Control

  • Under uncapped supply, owner control is measured as the owners’ currency balance divided by currency in circulation.Most interview participants considered fractional measurement appropriate, while one suggested relating owner-held currency to attainable hash power.Source: §4.1.1 · p. 13
  • High owner control can concentrate Proof-of-Stake consensus power and create a double-spending threat if an owner exceeds 50% of that power.The review cites Ethereum as an example of wealth concentration associated with pre-mined cryptocurrency.Source: §4.1.1 · p. 13

4.1.2 Improvement Protocol

  • The review describes measuring improvement-protocol control by examining moderation and proposal authorship, outcomes, and discussion.Azouvi et al. (2018) suggest using centrality metrics on proposal and comment data; IQMean is presented as a way to reduce sensitivity to outliers.Source: §4.1.2 · p. 14
  • Control over improvement protocols can shape network changes, while moderation by a subset can enable changes that lack broad community support.The review identifies the Bitcoin block-size debate, Bitcoin transaction-fee reduction, and Ethereum’s DAO-related hard fork as examples.Source: §4.1.2 · pp. 13–14

4.2 Network

The review identifies network centralization concerns in topology, geographic distribution, bandwidth, and routing. It describes measurement approaches and potential security, privacy, participation, and geopolitical implications, while noting gaps in measurement and real-world evidence.

4.2 Network

  • The review identifies topology formation, including node discovery and routing, as sources of network-layer centralization.It also describes connectivity as relevant to mining and block propagation.Source: §4.2 · p. 14

4.2.1 Node Discovery Protocol Control

  • No reviewed measurement technique specifically targets node discovery, although seed DNS nodes are identified as a centralization source.Centralized DNS can create a single point of failure and, if changed maliciously, enable eclipse attacks.Source: §4.2.1 · p. 15

4.2.2 Geographic distribution

  • The review reports a trend toward geographic concentration of nodes in Bitcoin and Ethereum, measured using peer-to-peer latency.The review links geographic concentration to potential geopolitical manipulation and faster propagation advantages among nearby nodes.Source: §4.2.2 · p. 15

4.2.3 Bandwidth Concentration

  • High bandwidth requirements may concentrate mining in data centers and limit consensus participation to entities with strong network capabilities.The review describes estimating peer bandwidth by measuring data-transfer time and grouping peers by provisioned bandwidth.Source: §4.2.3 · p. 16

4.2.4 Routing Centralization

  • The review reports that AS-level concentration has been studied for privacy and security, but finds no real-world incident reports in its sample.A network traversal that collects peer IP addresses is described for identifying AS-level structure.Source: §4.2.4 · p. 16

4.3 Consensus

Consensus centralization concerns the distribution of computational power, which supports Proof-of-Work agreement and affects security. The literature measures this distribution through block attribution and inequality measures, while identifying concentration-related attacks as a key implication.

4.3 Consensus

  • Proof-of-Work consensus power is miners’ hash power, and its concentration is a direct signifier of centralization.Eight interviewees raised consensus-power distribution unprompted as a prevalent concern.Source: §4.3 · p. 16, §4.3.1 · p. 16

4.3.1 Consensus Power Distribution

  • Commercial mining pools centralize control when pool managers choose included transactions and allocate work, although pooled mining itself is not considered a threat.The manager’s role requires participants to trust that manager.Source: §4.3.1 · p. 17
  • Block attribution identifies mining pools from observed blocks and coinbase addresses, then estimates each pool’s share of total computational power.Studies express the distribution as percentages or Gini values based on the Lorenz curve.Source: §4.3.1 · p. 17
  • The Gini coefficient measures consensus-power inequality from the Lorenz curve, ranging from 0 for complete equality to 1 for complete inequality.A is the area between the equality line and Lorenz curve; B is the area under the equality line.Source: §4.3.1 · p. 17
  • 26% consensus-power concentration can enable selfish mining, while concentration over 51% can result in a 51% attack.Smaller cryptocurrencies are more prone to 51% attacks; reported attacks averaged $2.5 million in losses per cryptocurrency.Source: §4.3.1 · p. 18

4.4 Incentive Layer

The incentive layer links cryptocurrency rewards and wealth distribution to public-blockchain security. The review discusses wealth concentration as a centralization concern, its measurement, and security and economic implications.

4.4 Incentive Layer

  • A fall in Bitcoin’s exchange rate below mining profitability may reduce hashing power, making it easier for attackers to gain consensus share and increasing attack threats.The review also identifies wealth concentration as a potential source of harm and discusses the Whale Transaction Attack.Source: §4.4 · p. 18
  • 60% of interview participants raised wealth concentration unprompted, with one describing centralization by tracing funds to a person or group.This framing applies regardless of the number of addresses involved.Source: §4.4 · p. 18

4.4.1 Wealth Concentration

  • In Bitcoin’s transaction graph, already wealthy nodes tend to increase their wealth at a higher speed than smaller nodes, a phenomenon Kondor et al. (2014) [103] call the “rich get richer” scheme.Source: §4.4.1 · p. 18
  • Wealth concentration can be measured as inequality in how much wealth a population controls, but address-level Gini calculations face complications from wallets that generate new addresses.One proposed response is to set a minimum balance for including an address; a simple percentage measure does not capture distribution.Source: §4.4.1 · p. 19
  • The review links wealth concentration to possible 51% and whale attacks, but notes that these potential attacks lacked real-world incident reports.Both are presented as potential security threats, including a 51% attack by a wealthy attacker during an exchange-rate fall.Source: §4.4.1 · p. 19
  • The iFish attack artificially inflated Ethereum transaction fees by 35%, while the bZx hack exploited a lending contract through high-value transactions and platform manipulation [102, 107].High wealth concentration may also enable attackers to raise transaction costs.Source: §4.4.1 · pp. 18–19

4.5 Operational Layer

The operational layer concerns participation costs, including mining equipment and blockchain storage. The review identifies specialized equipment and growing storage requirements as centralization concerns, with different measurement and participation implications.

4.5 Operational Layer

  • Specialized ASIC equipment raises miners’ capital and operating costs and creates a significant entry barrier in Bitcoin.The review classifies this as Specialized Equipment Concentration.Source: §4.5 · p. 19
  • High storage requirements can exclude traditional computing devices from consensus, concentrating participation among those able to afford greater resources.The paper notes that blockchain instances vary in storage demand, with Ethereum’s growth potentially limiting participation and increasing DDoS susceptibility as node numbers fall.Source: §4.5 · p. 19, §4.5.1 · p. 20

4.5.1 Size of the Blockchain

  • 20% of interview participants raised storage requirements; one participant described full-node operation becoming more expensive as capacity needs grow.The same interviewee suggested comparing storage growth with Moore’s law for hard drives.Source: §4.5.1 · pp. 19–20
  • Storage-growth rate is proposed as a way to capture storage-oriented centralization, using historical blockchain size data [114].The rate may be calculated periodically, ideally after each difficulty recalibration.Source: §4.5.1 · p. 20
  • Ethereum’s storage growth may limit participation, whereas Bitcoin is described as having no significant storage issue at the time discussed.Ethereum nodes are expected to store smart-contract code; fewer participating nodes increase the likelihood of a successful DDoS attack.Source: §4.5.1 · p. 20

4.5.2 Specialized Equipment Concentration

  • 60% of interview participants acknowledged specialized equipment concentration as an issue for public blockchains.Source: §4.5.2 · p. 20
  • The review reports no existing metric for hardware centralization and suggests nonpublic, often unavailable specialized hardware may explain the gap.Source: §4.5.2 · p. 20
  • Specialized equipment may advantage commercial entities, which could hoard efficient hardware and later release it when less profitable to operate.The review calls for further investigation of hardware measurement and its centralization impact; high entry barriers and old-hardware unprofitability have also been linked to greater likelihood of selfish-mining and double-spending attacks.Source: §4.5.2 · pp. 20–21

4.6 Application Layer

The application layer contains centralization risks in reference-client development, exchanges, and wallets. These risks can affect governance, security, and participation, while wallet concentration lacks a direct proposed measurement technique.

4.6.1 Reference Client Development Concentration

  • A few developers driving the reference client may gain influence over implementation changes and feed into blockchain governance concerns.The review distinguishes reference-client development from protocol-improvement centralization; it also notes that concentrated development may reflect highly skilled developers making useful contributions.Source: §4.6.1 · pp. 21–22

4.6.2 Exchange

  • Centralized exchanges can be single points of failure, with attacks exploiting exchange vulnerabilities and Mt. Gox’s closure following security flaws.Exchanges also facilitate price discovery, so the paper argues they should not be monopolized.Source: §4.6.2 · p. 22

4.6.3 Wallet Concentration

  • The literature reviewed offers no direct wallet-concentration measurement; exchange centralization may serve only as a rough proxy.The paper suggests the measurement gap may reflect wallets’ closed commercial environment.Source: §4.6.3 · p. 22
  • Centralized wallets may concentrate wealth, create dependence on their operators, and expose hosted-wallet services to targeted attacks.The review also describes wallets as potential single points of failure and a barrier to ecosystem entry.Source: §4.6.3 · p. 22

5 State of centralization in Bitcoin and Ethereum

The section surveys evidence on centralization across Bitcoin and Ethereum using the paper’s taxonomy. The reported evidence spans governance, network, consensus, wealth, operational requirements, and application-layer services.

5.1 Governance Layer

  • Bitcoin’s owner-control value is 0.033, compared with Ethereum’s 0.11; studies also report concentrated protocol-improvement contributions in both systems.The section describes a unilateral Bitcoin transaction-fee change and reports that a large share of Ethereum improvement proposals came from one contributor.Source: §5.1 · p. 23

5.2 Network Layer

  • Bitcoin is reported as more geographically centralized than Ethereum, whose average peer-to-peer latency is 26.7% higher; Bitcoin nodes also have about 1.9 to 2.7 times more bandwidth.The section additionally reports concentrated routing: 30% of Bitcoin’s network involved 10 ASes, while 28% of Ethereum nodes belonged to one AS.Source: §5.2 · pp. 23–24

5.3 Consensus Layer

  • Bitcoin’s top four mining pools controlled 53% of hashing power on average, while Ethereum’s top three controlled 61%, making Bitcoin less centralized by this comparison.Earlier Bitcoin estimates also report eight pools controlling 30% of hashing power in 2011 and one pool nearing 40% in 2014.Source: §5.3 · p. 24

5.4 Incentive Layer

  • Reported wealth measures indicate concentration in both systems: Bitcoin’s 2018 Gini value was 0.65, while Ethereum’s was 0.76.The Bitcoin estimate used a 185 BTC per-account threshold; the Ethereum estimate used 2477 ETH per account.Source: §5.4 · p. 24

5.5 Operational Layer

  • A Bitcoin full node was reported to require 204 GB of storage, compared with 385 GB for Ethereum; Bitcoin storage grew about 0.1-0.5 GB per day.The review found no longitudinal studies tracking storage requirements over a long period and no proposed measurement techniques for specialized-equipment concentration.Source: §5.5 · p. 24

5.6 Application Layer

  • In Bitcoin, seven exchanges served more than 95% of trades in 2015, and 18 of 40 examined exchanges had closed, wiping out customers’ account balances.The section reports no studies explicitly describing Ethereum exchange behavior and no suggested measurement technique for wallet centralization.Source: §5.6 · p. 25
  • The section characterizes Bitcoin’s principal threats as network, consensus, and application-layer centralization, while Ethereum’s include governance, consensus, incentive, and application layers.For Bitcoin, the named network concerns include geographic distribution, bandwidth, and routing; for Ethereum, the section highlights unilateral governance decisions and wealth concentration.Source: §5.6 · p. 25

6 Discussion

The discussion argues that public-blockchain centralization is multidimensional rather than binary, and that its implications differ across aspects and blockchain instances. The taxonomy supports structured analysis while the authors call for empirically grounded assessment of each aspect.

6.1 Non Binary Nature of Centralization

  • Centralization is defined as architectural dimensions restricting most participants through direct or indirect economic, social, or technical constraints, rather than as a binary state.The authors report that decentralization is loosely defined and often treated as binary in literature not focused specifically on decentralization.Source: §6.1 · p. 25
  • The taxonomy offers a structured basis for discussing centralization, while a more granular definition may dilute meaning for non-experts.One interviewee contrasted engineers’ preference for precision and multidimensional models with business audiences’ interest in a single score.Source: §6.1 · p. 26
  • Consensus-power concentration was the most recognized concern, and Bitcoin’s top four pools controlled over 53% of hashing power versus over 61% for Ethereum’s top three.The authors connect this concern to security threats such as double spending and selfish mining.Source: §6.1 · p. 26
  • Specialized, often proprietary, hardware can indirectly limit participation because profitable mining may require efficient equipment.The paper notes proposals for memory-heavy consensus algorithms, for which specialized hardware design is inefficient.Source: §6.1 · p. 26
  • Bitcoin storage grows around 0.1 to 0.5 GB per day, and increasing requirements may prevent conventional devices from serving as nodes.The authors suggest further investigation of storage optimization.Source: §6.1 · p. 26
  • Half of interviewees accepted node-discovery protocol control as a decentralization threat, despite only one reviewed article reporting on it.The authors suggest that DNS-node traffic may delay new-node connections, a practical effect not often accounted for in network simulations.Source: §6.1 · p. 26
  • The authors report that Ethereum is more geographically spread out than Bitcoin, but their review identified no strategies to address geographic concentration.They reason that Ethereum’s use of conventional hardware such as GPUs may contribute to this difference.Source: §6.1 · p. 26
  • Centralized wallet and exchange services may advantage their operators through stored cryptocurrency; seven exchanges were reported to serve over 95% of Bitcoin trades.The discussion also notes that wealth concentration is recognized by interviewees and literature, but most reported research concerns Proof-of-Work rather than Proof-of-Stake.Source: §6.1 · p. 26

7 Conclusion

The review synthesizes centralization research into an initial taxonomy and reports similarities and differences between Bitcoin and Ethereum. It argues that centralization should be evaluated aspect by aspect, while noting limits of the review and areas for further study.

7.1 Contribution

  • The initial taxonomy organizes diverse centralization concepts and measurements around a generic blockchain architecture.The review also identifies adverse impacts, including security threats, and a gap in Ethereum-specific centralization research.Source: §7.1 · p. 27, §7 · p. 27
  • Ethereum is reported as more centralized than Bitcoin in wealth and consensus power, while its nodes are more geographically dispersed.The two platforms have similar issues involving reference clients, protocol development, and exchanges.Source: §7.1 · pp. 27–28
  • Centralization is not necessarily adverse in every aspect; the authors recommend assessing its impact empirically for each aspect.They propose moving beyond a binary view toward a multidimensional scale for measurement and control where needed.Source: §7.1 · p. 28

7.2 Threats to validity

  • The review may have missed relevant articles with different phrasing or absent from its repositories; Google Scholar coverage stopped at its top 1,000 results.Relevant articles dropped off significantly after the top 400 results.Source: §7.2 · p. 28
  • Strict title-and-abstract filtering and the inclusion of articles not primarily focused on centralization may constrain the shortlisted literature.The authors report quality review of 212 shortlisted articles, yielding a final set of 89.Source: §7.2 · p. 28
  • Interviews relying on contemporaneous notes may have missed some observations, despite transcript checks and interviewee validation..Source: §7.2 · p. 28

7.3 Future Work

  • The authors propose case studies of individual cryptocurrencies and implementations, and research into potential solutions to identified centralization avenues.They also suggest a single objective centralization score as a possible aid for end-users and nonspecialist researchers.Source: §7.3 · pp. 28–29
  • Future work could develop implementation-specific taxonomies and examine previously unidentified centralization factors.The current generic taxonomy omits consensus-specific issues such as Stake bleeding and source-code dependencies in smart contracts.Source: §7.3 · pp. 28–29

Related papers

Loading 2009.12542v1…