Pith. sign in

REVIEW 3 major objections 4 minor 23 references

Agent-based Simulation of Blockchains

T0 review · 3 major / 4 minor · reviewed 2026-08-14 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read The paper presents LUNES-Blockchain, an agent-based Bitcoin simulator built on parallel and distributed simulation, and demonstrates it on a 10,000-node Sybil attack scenario.

desk verdict A small, honest tool paper whose main selling point—PADS scalability—is inherited, not demonstrated; the attack results are a credible percolation exercise. read the letter →

arxiv 1908.11811 v2 pith:JHXRHEEH submitted 2019-08-29 cs.CR cs.DCcs.MA

Edoardo Rosa , Gabriele D'Angelo , Stefano Ferretti This is my paper · ORCID
classification cs.CRcs.DCcs.MA
keywords blockchainsimulationagent-basedparallelanddistributedBitcoinSybilattackdenialofservicepeer-to-peernetworksLUNES-Blockchain
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

LUNES-Blockchain is an agent-based blockchain simulator built on the parallel-and-distributed simulation (PADS) substrate offered by ARTIS/GAIA, and the paper's central claim is that it is the first blockchain simulator able to exploit the speedup and scalability that PADS provides. To support that claim, the authors implement a simplified Bitcoin model with 10,000 nodes, 70% miners, gossip-based dissemination, and a difficulty-aware but computationally cheap Proof-of-Work mining process. They then study a Sybil-style filtering denial-of-service attack in which malicious agents silently drop every message originating from a victim node. The measured effect is a steady decline in the victim's reach as the number of attackers grows, a sharp drop beyond 7,000 malicious nodes, and complete isolation above 7,500, which the authors take as consistent with known theoretical expectations. The paper's point is that a full blockchain stack—overlay, dissemination, consensus, and attack behaviour—can be expressed agent-by-agent on a PADS platform, making blockchain simulation a candidate for parallel and distributed execution.

What carries the argument

The central mechanism is the one-agent-per-node mapping: every blockchain node becomes a Simulated Entity in the GAIA multi-agent layer, and these agents interact only through timestamped messages, so protocol behaviour is emergent rather than scripted globally. ARTIS/GAIA then clusters frequently communicating agents into the same Logical Process, which is the device that is supposed to contain inter-process communication and make parallel or distributed execution efficient. LUNES contributes the degree-dependent dissemination protocol used to model transaction and block propagation, while LUNES-Blockchain adds the blockchain-specific agent states and actions: transaction generation, a difficulty- and hashrate-based mining model that avoids the cost of real hash computations, and the time-stepped clock calibrated to Bitcoin. The Sybil attack is itself implemented as an agent behaviour—malicious agents drop all messages originating from the victim—which means the attack's reach curve is an emergent property of many local decisions.

What would settle it

Running the published 10,000-node Sybil scenario on 2, 4, 8, and 16 Logical Processes across separate cores and comparing wall-clock time and maximum network size against the reported 60-second single-LP baseline would settle whether PADS delivers the speedup and scalability that define the contribution.

Watch

Extended reading notes

Core claim

The core claim, stated as directly as the authors state it, is that LUNES-Blockchain is the first blockchain simulator able to take advantage of the performance speedup and extended scalability offered by Parallel and Distributed Simulation. The simulator represents each blockchain node as an autonomous agent that keeps a local ledger, generates and validates transactions, mines new blocks at a rate determined by its hashrate and the global difficulty, and broadcasts information through a degree-dependent gossip protocol supplied by the LUNES layer. A time-stepped clock maps one simulated minute per step, so on average a block appears every ten steps, matching Bitcoin's cadence. The demonstration scenario is a filtering DoS attack on a 10,000-node network: malicious nodes drop all messages from a chosen victim, and the simulator measures how many nodes still receive the victim's messages as the number of attackers ranges from 1 to 9,999. The reported curves show reach declining with attacker count, a sharp fall when attackers exceed 7,000, and total disconnection of the victim beyond 7,500, which the authors describe as comparable to theoretical expectations for Sybil attacks against peer-to-peer botnets and as confirmation that agent-based blockchain modelling on PADS is viable.

Load-bearing premise

The load-bearing premise is that the underlying parallel-and-distributed simulation substrate delivers the speedup and scalability LUNES-Blockchain is designed for, since every experiment reported here runs sequentially on a single Logical Process and no parallel or distributed run is shown.

Editorial extensions

If this is right

  • Under this design, blockchain anomaly and attack scenarios can be studied at network scales that sequential simulators struggle with, because the platform is built to spread Logical Processes across processing units.
  • The measured threshold—full isolation of a victim only when roughly three-quarters of the ten-thousand-node network attacks—gives a concrete, topology-dependent baseline for how many malicious nodes a filtering defence must tolerate.
  • Because mining is modelled through hashrate and difficulty rather than real cryptography, experiments on block interval, difficulty, and pool behaviour become cheap enough to sweep across many configurations.
  • The same agent-based stack is extensible to other consensus protocols and to smart-contract blockchains; the authors identify Ethereum-style modelling as the planned next step.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • An implication the paper leaves implicit is that if the PADS speedup materializes, the limiting factor for blockchain simulation shifts from event throughput to synchronization and inter-Logical-Process communication, making topology-aware agent partitioning at least as important as raw node count.
  • The sharp drop near 7,000 attackers looks like a percolation transition in a 10,000-node random graph with 8 edges per node; computing the graph's connectivity threshold would separate purely structural effects from the Bitcoin-specific protocol effects.
  • A direct testable extension would vary average node degree and graph topology while holding the attack fixed: if the 7,000 and 7,500 thresholds move, then Sybil isolation is governed mainly by graph connectivity, not by the consensus layer.
  • If the planned Dandelion/Dandelion++ module is added, the simulator could compare how the anonymity phase and the flooding phase each degrade under filtering attacks, a comparison no current simulator seems to offer.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 4 minor

Summary. This paper presents LUNES-Blockchain, an agent-based blockchain simulator built on the ARTIS/GAIA parallel and distributed simulation middleware and the LUNES network simulator. The authors describe a simplified Bitcoin model, set parameters to mimic the Bitcoin network in Q4 2018, and study a filtering denial-of-service attack in which a set of malicious nodes drops all messages originating from a target node. The main claims are that LUNES-Blockchain is the first blockchain simulator to exploit the performance speedup and extended scalability provided by PADS, and that the reported results confirm the viability of agent-based blockchain modelling via PADS. All reported experiments use a single Logical Process in sequential mode, and the paper explicitly states that the work is preliminary and does not represent a validation.

Significance. If the PADS scalability claim were backed by experiments, this would be a useful contribution: it would provide an agent-based blockchain simulator with a potential path to large-scale parallel and distributed simulation, and the security case study would illustrate how such a simulator can be used to explore attacks. The paper is transparent about its preliminary status, provides a parameter table (Table 1) to support reproducibility, and makes the simulator available for peer review. However, the central PADS claim is currently unmeasured, and the attack evaluation lacks a quantitative comparison to theory, so the current contribution is an architectural proposal with a small-scale sequential case study rather than a demonstration of the claimed capabilities.

major comments (3)
  1. [Section 4, first paragraph; Conclusions] The central claim of the paper—that LUNES-Blockchain is able to exploit the performance speedup and extended scalability provided by PADS—is not supported by the reported experiments. The paper states that all runs used "a single LP (i.e. sequential simulation)" and provides no multi-LP, multi-core, or multi-host results. No wall-clock time, memory usage, or speedup is reported for any parallel configuration, and the only performance datum is a 60-second sequential run for 10,000 nodes. Consequently, the conclusions in Section 5 and the Abstract overstate what the data show; at most, the paper demonstrates that the existing ARTIS/GAIA middleware is architecturally compatible with the blockchain model. Please add at least one experiment that varies the number of LPs/PEUs and reports speedup and scalability relative to the sequential baseline.
  2. [Section 4, paragraph around Figures 2 and 3] The evaluation of the DoS attack claims that the outcomes are "comparable with theoretical results expected for this kind of attack against peer-to-peer botnets" and cites Ref. [21], but no theoretical curve, error bars, confidence intervals, or statistical test is provided. The paper also explicitly states that this "does not represent a validation". Without a quantitative comparison, the results in Figures 2 and 3 do not substantiate the abstract's assertion that the results "confirm the viability" of the simulator. Please specify the theoretical benchmark, show it in the figures or in a table, and report the deviation between simulation and theory.
  3. [Section 4, first paragraph of the attack description] The attack is described as "a Sybil Attack", but the implemented mechanism—a set of malicious nodes silently dropping all messages from a targeted node, leading to complete isolation when the target is fully surrounded—corresponds to an eclipse attack or a filtering denial-of-service, not to a Sybil attack in the standard sense (where an adversary forges multiple identities). The paper does not model identity creation or the adversarial control of multiple distinct nodes. Using the term "Sybil" is misleading and affects how the results should be interpreted; please rename the attack or justify the terminology with a suitable reference.
minor comments (4)
  1. [Section 3.3, Table 1] The model uses a random graph topology and a degree-dependent gossip protocol in place of Bitcoin's actual peer-to-peer topology and Dandelion/Dandelion++ dissemination, which the paper acknowledges. However, the paper also claims to mimic Bitcoin's mining difficulty and hashrate; it should provide evidence that the simulated block interval is centered near the target 10 minutes (10 time-steps) and note how the simplified topology affects dissemination delays.
  2. [Section 3.3, paragraph on time modeling] There is a typo: "the simulated time is dived in a sequence of time-steps" should be "divided". In addition, in the same section, "assess if it possible" should be "assess if it is possible". A duplicate "the the" appears in Section 3 before Section 3.1.
  3. [Section 4, Figure 3] The caption of Figure 3 says "Different setups with an increasing number of attacking nodes", but the figure shows curves for different attack configurations without a legend or explicit mapping from line style to attacker count. Please add a legend or annotate the curves.
  4. [Section 4, first paragraph after Table 1] The paper reports that a single run takes "an average of 60 seconds", but does not report the number of runs used to compute the average or the variance. Please provide these details to support the performance statement.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: the attack results are not fitted to model parameters, and the central PADS scalability claim is an unmeasured extrapolation rather than a derivation that reduces to its own inputs.

full rationale

The paper's derived results are the Sybil/filtering DoS outcomes in Section 4. The parameters in Table 1 are fixed values meant to mimic the Bitcoin network as of Q4 2018; they are not fitted to the attack curves. The measured quantity (average number of nodes reached by messages from the attacked node) is produced by the simulation model, not imposed by construction, and the comparison to peer-to-peer botnet expectations in [21] is explicitly hedged: "Clearly, this does not represent a validation of the proposed simulation model but it is a positive outcome." Thus the reported experiment is not a fitted input renamed as a prediction. The model does inherit ARTIS/GAIA and LUNES from the authors' prior work, and this self-citation is load-bearing for the claim that LUNES-Blockchain is "able to take advantage of the performance speedup and extended scalability provided by PADS." However, the paper's own execution data state that each run used "a single LP (i.e. sequential simulation)", so no PADS speedup is actually measured. That is a missing-evidence/correctness gap, not circularity: there is no equation or construction in which the central claim is equivalent to its input. The paper even flags the limitation by promising "a full-fledged validation and scalability assessment" in future work. Given one substantive reliance on self-cited infrastructure but no demonstrated reduction of the central claim to its own inputs, the appropriate circularity score is 1.

Assumptions & free parameters 9 free parameters · 6 assumptions · 0 invented entities

The central claim rests on a set of manually chosen Bitcoin-mimicking parameters and several domain assumptions about network topology and dissemination fidelity. The parameters are not fitted to the attack output, but they determine the quantitative results. The most consequential untested assumption is that ARTIS/GAIA will provide the claimed parallel scalability, since no PADS execution is reported.

free parameters (9)
  • TTL = 16
    Gossip message time-to-live, set to 16, chosen to mimic Bitcoin-like flooding but not fitted to the attack results.
  • Dissemination probability function = degree-dependent, coefficients 4 and 74
    Parameters of the degree-dependent gossip function taken from prior LUNES work [6], not fitted to the attack results.
  • END_CLOCK = 5000 time-steps
    Duration of each simulation run, representing 5000 simulated minutes.
  • NODES = 10000
    Network size, based on estimated Bitcoin node count around 10,400 [3].
  • MINERS COUNT = 70%
    Fraction of nodes acting as miners, a modeling choice.
  • DIFFICULTY = 6489747252517
    Bitcoin difficulty value from Q4 2018, used in mining probability.
  • HASHRATE = 43983561622000000000
    Total network hashrate from Q4 2018, used to scale mining rates.
  • Edges per Node = 8
    Average degree of the random graph overlay, a modeling choice affecting dissemination and attack impact.
  • Time-step length = 1 minute
    The mapping from simulation time-step to real time, used to align block interval to 10 minutes.
assumptions (6)
  • domain assumption Random graph with 8 edges per node is a sufficient model of the Bitcoin P2P overlay.
    Section 3.3 and Table 1. The paper acknowledges a more accurate topology is future work [19].
  • domain assumption Degree-dependent gossip dissemination approximates Bitcoin's Dandelion/Dandelion++ protocol.
    Section 3.3 states Dandelion is under development and a LUNES gossip protocol is used instead.
  • domain assumption Each mining pool is treated as a single node.
    Section 3.3, justified for the specific malicious behavior studied.
  • domain assumption Time-stepped synchronization with 1-minute steps preserves block and transaction propagation dynamics.
    Section 3.3 states each step is one minute and a block is created every 10 steps on average.
  • ad hoc to paper Filtering DoS that drops all messages from a given node constitutes a Sybil attack.
    Section 4. This is a nonstandard use of the term Sybil; a Sybil attack normally involves creating fake identities. This assumption shapes the experiment.
  • domain assumption ARTIS/GAIA's PADS services yield speedup and scalability for this model.
    Sections 3.1 and 4. The paper claims PADS benefits but only reports a sequential (single LP) run.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Agent-based Simulation of Blockchains." pith.science (2026). https://pith.science/paper/JHXRHEEH

@misc{pith2026190811811,
  author       = {Pith},
  title        = {Pith review of: Agent-based Simulation of Blockchains},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/JHXRHEEH}},
  note         = {Machine review of arXiv:1908.11811}
}
read the original abstract

In this paper, we describe LUNES-Blockchain, an agent-based simulator of blockchains that is able to exploit Parallel and Distributed Simulation (PADS) techniques to offer a high level of scalability. To assess the preliminary implementation of our simulator, we provide a simplified modelling of the Bitcoin protocol and we study the effect of a security attack on the consensus protocol in which a set of malicious nodes implements a filtering denial of service (i.e. Sybil Attack). The results confirm the viability of the agent-based modelling of blockchains implemented by means of PADS.

Figures

Figures reproduced from arXiv: 1908.11811 by the authors.

Figure 1
Figure 1. Blockchain protocol stack. analysis of a DoS attack on the Bitcoin network is described in Section 4. Finally, Section 5 provides some concluding remarks. 2. Background and Related Work Blockchain is a technology that was initially proposed in the Bitcoin system, in 2009, by an anonymous author with the pseudo-name of Satoshi Nakamoto [18]. In its first wave, the blockchain allowed parties to transact directly, i.e.… view at source ↗
Figure 2
Figure 2. Average number of reached nodes during a DoS filtering attack. [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Number of reached nodes for each transaction during a DoS filtering attack. Different setups with an increasing [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

23 extracted references · 22 canonical work pages

  1. [21]

    A. L. Verigin. Evaluating the Effectiveness of Sybil Attacks Against Peer-to-Peer Botnets. http://hdl.handle.net/1828/5095, 2018

    work page 2018

  2. [1]

    Alharby and A

    M. Alharby and A. van Moorsel. Blocksim: A simulation frame- work for blockchain systems. SIGMETRICS Perform. Eval. Rev. , 46(3):135–138, Jan. 2019

    work page 2019

  3. [2]

    Y . Aoki, K. Otsuki, T. Kaneko, R. Banno, and K. Shudo. Simblock: A blockchain network simulator. In Proc. of the 2nd Workshop on Cryp- tocurrencies and Blockchains for Distributed Systems , CryBlock’19. IEEE, 2019

    work page 2019

  4. [3]

    Global Bitcoin Nodes Distribution

    Bitnodes. Global Bitcoin Nodes Distribution. https://bitnodes.earn. com/, 2019

    work page 2019

  5. [4]

    V . Buterin. A next-generation smart contract and decentralized application platform. White Paper, 2018. https://github.com/ethereum/ wiki/wiki/White-Paper, Accessed on 2018-03-02

    work page 2018

  6. [5]

    Castro and B

    M. Castro and B. Liskov. Practical byzantine fault tolerance. In Proceedings of the Third Symposium on Operating Systems Design and Implementation, OSDI ’99, pages 173–186, Berkeley, CA, USA,

    work page

  7. [6]

    D’Angelo and S

    G. D’Angelo and S. Ferretti. Highly intensive data dissemination in complex networks. Journal of Parallel and Distributed Computing , 99:28 – 50, 2017

    work page 2017

  8. [7]

    D’Angelo and S

    G. D’Angelo and S. Ferretti. Parallel And Distributed Simulation (PADS) Research Group. http://pads.cs.unibo.it, 2019

    work page 2019

Show all 23 references
  1. [8]

    D’Angelo, S

    G. D’Angelo, S. Ferretti, and M. Marzolla. A blockchain-based flight data recorder for cloud accountability. In Proc. of the 1st Workshop on Cryptocurrencies and Blockchains for Distributed Systems , Cry- Block’18, pages 93–98, New York, NY , USA, 2018. ACM

    2018

  2. [9]

    G. DAngelo. The simulation model partitioning problem: an adaptive solution based on self-clustering. Simulation Modelling Practice and Theory (SIMPAT), 70:1 – 20, 2017

    2017

  3. [10]

    Egea-Lopez, J

    E. Egea-Lopez, J. Vales-Alonso, A. Martinez-Sala, P. Pavon-Mario, and J. Garcia-Haro. Simulation scalability issues in wireless sensor networks. Communications Magazine, IEEE , 44(7):64 – 73, july 2006

    2006

  4. [11]

    Eyal and E

    I. Eyal and E. G. Sirer. Majority is not enough: Bitcoin mining is vulnerable. Commun. ACM, 61(7):95–102, June 2018

    2018

  5. [12]

    Ferretti

    S. Ferretti. Gossiping for resource discovering: An analysis based on complex network theory. Future Generation Computer Systems , 29(6):1631 – 1644, 2013

    2013

  6. [13]

    Fujimoto

    R. Fujimoto. Parallel and Distributed Simulation Systems . Wiley & Sons, 2000

    2000

  7. [14]

    A. E. Gencer, S. Basu, I. Eyal, R. van Renesse, and E. G. Sirer. Decen- tralization in bitcoin and ethereum networks. CoRR, abs/1801.03998, 2018

    2018 arXiv

  8. [15]

    Gervais, G

    A. Gervais, G. O. Karame, K. W ¨ust, V . Glykantzis, H. Ritzdorf, and S. Capkun. On the security and performance of proof of work blockchains. In Proceedings of the 2016 ACM SIGSAC Conference on Computer and Communications Security , CCS ’16, pages 3–16, New York, NY , USA, 2016. ACM

    2016

  9. [16]

    Maymounkov and D

    P. Maymounkov and D. Mazi `eres. Kademlia: A peer-to-peer infor- mation system based on the xor metric. In P. Druschel, F. Kaashoek, and A. Rowstron, editors, Peer-to-Peer Systems, pages 53–65, Berlin, Heidelberg, 2002. Springer Berlin Heidelberg

    2002

  10. [17]

    Miller and R

    A. Miller and R. Jansen. Shadow-bitcoin: Scalable simulation via direct execution of multi-threaded applications. In 8th Workshop on Cyber Security Experimentation and Test (CSET 15) , Washington, D.C., Aug. 2015. USENIX Association

    2015

  11. [18]

    Nakamoto

    S. Nakamoto. Bitcoin: A peer-to-peer electronic cash system, http://bitcoin.org/bitcoin.pdf, 2009

    2009

  12. [19]

    Neudecker, P

    T. Neudecker, P. Andelfinger, and H. Hartenstein. Timing analysis for inferring the topology of the bitcoin peer-to-peer network. In 2016 Intl IEEE Conferences on Ubiquitous Intelligence Computing, Advanced and Trusted Computing, Scalable Computing and Communications, Cloud and...

    2016

  13. [20]

    Stoykov, K

    L. Stoykov, K. Zhang, and H.-A. Jacobsen. Vibes: Fast blockchain simulations for large-scale peer-to-peer networks: Demo. In Proceed- ings of the 18th ACM/IFIP/USENIX Middleware Conference: Posters and Demos , Middleware ’17, pages 19–20, New York, NY , USA,

  14. [22]

    Y . Xiao, N. Zhang, W. Lou, and Y . T. Hou. A survey of distributed consensus protocols for blockchain networks. CoRR, abs/1904.04098, 2019

    1904 arXiv

  15. [23]

    Zichichi, M

    M. Zichichi, M. Contu, S. Ferretti, and G. D’Angelo. Likestarter: a smart-contract based social dao for crowdfunding. In Proc. of the 2nd Workshop on Cryptocurrencies and Blockchains for Distributed Systems, CryBlock’19. IEEE, 2019

    2019

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.