{"id":"c34414a7-aa5f-4a8f-8a5b-3d2f9b4b6209","arxiv_id":"2506.08636","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A large systematic review of 921 papers finds four research patterns in blockchain-edge computing work and shows blockchain is mostly used to secure edge systems.","lead":"This paper reviews almost 1,000 research papers on combining blockchain and edge computing, sorting them into 22 categories. It finds four main research patterns and shows that most work uses blockchain to secure edge systems, especially for mobile computing.","discovery_kind":"review","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Inter-rater reliability of the 921-paper taxonomy coding is never reported; the percentages and MCA patterns rest entirely on unverified manual labels.","rationale":"The reader's weakest assumption correctly identifies the manual coding reliability as the central vulnerability of the paper. My reading of the full text confirms that the methods section describes the screening process but gives no agreement measure for the taxonomy coding, and the analysis pipeline (descriptive statistics, MCA, clustering) consumes the coded dataset as ground truth. The paper does provide the dataset and code openly, which is a real strength and makes an independent reliability check feasible, but it does not by itself validate the labels. The secondary concern about the unverified 2023–2025 trend confirmation is real but less central, since the core claim concerns the 2015–2022 corpus. The MCA methodology is appropriate for categorical data, but its output inherits any coding instability. I therefore agree with the CONDITIONAL verdict: the paper is a valuable and unusually large SLR, but the quantitative claims should not be accepted as final until coding reliability is demonstrated. No change from the reader's verdict is needed.","tokens_in":28697,"tokens_out":2048,"duration_ms":26427,"concrete_test":"Independently re-code a random sample of 100 papers from the final 921 (about 11%) using the taxonomy in Table II, with at least two coders who are blind to the original labels and to each other. Compute per-dimension Fleiss' kappa or Krippendorff's alpha for the key categorical dimensions (Permission, Type, Problem, Application, Contribution, Consensus, Chain, Reward). If kappa falls below 0.6 for any of these dimensions, the reported percentages and the four MCA patterns cannot be treated as stable; if kappa is at or above 0.8, the concern is substantially mitigated. As a complementary check, re-run the MCA on a bootstrap-resampled or noise-perturbed version of the full dataset and verify that the same four meta-dimensions emerge with stable attribute loadings.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central contribution is a quantitative map of the blockchain–edge computing nexus built from 921 papers coded along 22 dimensions with 287 attributes. Every reported percentage (e.g., 75% Perspective 1, 22% permissionless), every Figure 4–5 trend, and the four MCA patterns in Figure 6 are derived from this manual coding. Section IV-A describes two exclusion rounds with 'five reviewers independently assessing the exclusion criteria,' but no inter-rater reliability statistic (Cohen's kappa, Fleiss' kappa, Krippendorff's alpha) is reported for the exclusion decisions or for the taxonomy classification itself. The taxonomy in Table II includes dimensions with genuinely ambiguous boundaries, such as 'Problem: Security' vs. 'Problem: Privacy,' 'Contribution: Framework' vs. 'Architecture,' and TRL level assignments. Without evidence of coding stability, a different team could plausibly produce different labels, shifting the reported percentages and potentially altering which combinations of attributes dominate the MCA. The paper's assertion that papers published in the last three years 'confirm the key trends' is a bare statement with no citation or comparative data, so it cannot rescue the analysis. This is not a question of the review's scope or novelty; it is the load-bearing assumption that the manually coded dataset is reliable enough to support the quantitative and MCA conclusions.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This manuscript reports a systematic literature review of the blockchain-edge computing nexus, covering 5,903 papers retrieved from Scopus, ScienceDirect, and Web of Science, of which 921 were retained after screening. The authors construct a taxonomy with 22 dimensions and 287 attributes and use it to code the retained papers. They report descriptive statistics, temporal trends, and a Multiple Correspondence Analysis from which they derive four 'study patterns': private permissioned design, technology implementation, proof of concept research, and public permissionless design. The paper also states that blockchain-assisted edge computing for privacy and security is the dominant research direction. The authors make the coded dataset and analysis code publicly available.","tokens_in":28889,"tokens_out":4611,"duration_ms":51863,"significance":"If the underlying coded dataset is reliable, this is the largest systematic quantitative review of blockchain-edge computing to date, and it offers a reusable taxonomy and an open dataset that can support future meta-analyses. The three-perspective framing (blockchain-for-edge, edge-for-blockchain, synergistic) is a useful organizing device, and the attempt to use MCA to identify latent study patterns is methodologically distinctive in this literature. The public availability of the dataset and code is a clear strength and supports reproducibility. However, the paper's quantitative claims, including the reported percentages and the four MCA-based patterns, all rest on the manual assignment of 921 papers to 287 taxonomy attributes, and the manuscript provides no evidence of coding reliability. The significance of the contribution is therefore conditional on resolving this verification gap.","major_comments":[{"comment":"The central quantitative results (e.g., the 75%/19%/6% perspective distribution in Section V-A, the permissioned/permissionless percentages in Section V-B, and the four MCA patterns in Section V-C) are derived entirely from the authors' manual coding of 921 papers along the 22-dimension taxonomy. Although Section IV-A states that five reviewers independently assessed exclusion criteria, no inter-rater reliability statistic (Cohen's kappa, Fleiss' kappa, or Krippendorff's alpha) is reported for the exclusion decisions or for the taxonomy classification itself. Given that several dimension boundaries are genuinely ambiguous (e.g., Problem: Security vs. Problem: Privacy, Contribution: Framework vs. Architecture, and TRL assignments), the stability of the coded dataset is a load-bearing assumption. Please add a formal reliability assessment, for example by having multiple coders independently code a random subsample and reporting agreement statistics, or otherwise provide evidence that the coding is consistent across raters.","section":"Section IV-A and Section III"},{"comment":"The claim that 'four combinations are sufficient to capture the variance' is not supported by the reported information. Figure 6a displays eigenvalues on the order of 0.01-0.04, which in MCA typically corresponds to a small fraction of total inertia, yet no cumulative explained variance or eigenvalue-based selection criterion is reported. Without a stated threshold (e.g., cumulative inertia at some percentage, or a scree-test rule), the selection of the top four meta-dimensions and the resulting four patterns is not justified. Please report the cumulative variance accounted for by the selected dimensions and the criterion used to decide that four dimensions are sufficient.","section":"Section V-C and Figure 6a"},{"comment":"The statement that 'Other articles published in the last 3 years confirm the key trends captured within this chosen time period' is made without citation, data, or comparison. Because the paper was published in 2025 but the corpus ends in October 2022, the representativeness of the time window is central to the paper's claim to describe the 'current state' of the field. This assertion needs empirical support, for instance a supplementary search covering 2022-2025 with a comparison of key attribute distributions, or it should be removed and the claims explicitly scoped to 2015-October 2022.","section":"Section IV-A"}],"minor_comments":[{"comment":"The figure labels are inconsistent and confusing: the caption lists '(c-f) top 20 attributes contributed to the four meta-dimensions,' but the internal labels duplicate '(c) Top 20 attributes contributed to Meta-Dimension 1/2' and then switch to '(e) Central Traits...' and '(f) Central Traits...' for Meta-Dimensions 3 and 4. Please harmonize the subfigure labels with the pattern names used in the text.","section":"Figure 6"},{"comment":"There are several typographical inconsistencies in technology names: 'V ANET' in Table II, 'Resperry Pi' and 'Resberry Pi' in Figure 6, and 'Raspberry Pi' in the text. Please standardize these spellings.","section":"Table II and Figure 6"},{"comment":"Figure 1 lists an exclusion criterion 'Extended Paper: has been expanded or supplemented in another publication,' but this criterion is not described in the text of Section IV-A. Please either add a description of how this criterion was applied or remove it from the figure for consistency.","section":"Section IV-A and Figure 1"},{"comment":"The statement that 'the primary problem addressed in Perspective 1 is security' should be supported by a statistical test or at least a confidence measure, since Figure 3 shows security and performance percentages that are visually close. A difference-of-proportions test would make the claim more robust.","section":"Section V-A"}],"recommendation":"major_revision","confidential_remarks":"The open dataset and code are valuable assets, and the review is clearly a substantial piece of work. However, the central empirical claims hinge on the reliability of manual coding, and the absence of any inter-rater reliability measure is a significant gap that should be addressed before publication. The MCA variance-explained issue is also important because it directly affects the validity of the four claimed patterns. I believe these concerns are fixable within the scope of the manuscript, but they require additional analysis or reporting. The paper's fit with cs.DC is appropriate, and the self-citations are not excessive."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First thing to know: this is the largest systematic review of the blockchain–edge computing nexus I've seen, and the first to apply MCA to a coded corpus this size. It puts the dataset online, which is the right way to do this kind of work. The headline result—75% of papers take the blockchain-helps-edge perspective, permissioned designs dominate, security/privacy is the leading motivation—is plausible and roughly consistent with the qualitative surveys in the area.\n\nWhat it does well: the scale. 921 papers, 22 dimensions, 287 attributes, and a transparent PRISMA-style search. The taxonomy is fine-grained and mostly well defined. The MCA is a legitimate tool for this data, and the four patterns (private permissioned design, technology implementation, proof-of-concept research, public permissionless design) are readable and connect to the descriptive statistics. Table I is a fair comparison with prior reviews.\n\nThe soft spot is the one the stress-test flags: no inter-rater reliability for the coding. Five authors coded 921 papers independently in screening rounds, but no kappa is reported for the taxonomy classification. Some dimension boundaries are genuinely fuzzy—security vs. privacy, framework vs. architecture, TRL levels. This matters most for the MCA patterns, which can shift with label noise. I don't think it breaks the paper, because the dominant proportions are large and the qualitative conclusions are corroborated by prior reviews. But the absence of any reliability estimate is a real gap, and the authors should either compute one for a sample or add a sensitivity analysis.\n\nSecond issue: the claim in Section IV-A that papers from 2022–2025 'confirm the key trends' is a bare assertion, no citation or data. Given that the review stops in October 2022, this needs either evidence or deletion.\n\nMinor: some typos in the figures. Self-citation is present but not abused; the cited taxonomy paper [8] is a reasonable methodological basis.\n\nBottom line: this is a useful field map for anyone entering blockchain-edge research, and the open dataset is a genuine contribution. It deserves peer review. The reliability gap is addressable and the post-2022 claim is easy to fix. I'd send it to a serious venue with the expectation of a revision.","headline":"Largest coded map of the blockchain-edge literature to date, with open data—but the quantitative core lacks reported inter-rater reliability.","tokens_in":29469,"tokens_out":2888,"would_cite":true,"duration_ms":33648,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"The largest systematic map of blockchain and edge computing research to date, drawn from 921 papers, finds the field splits into four recurring study patterns.","keywords":["systematic literature review","edge computing","blockchain","distributed ledger","multiple correspondence analysis","taxonomy","permissioned vs permissionless","research patterns"],"falsifier":"Take a random sample of roughly 100 of the 921 papers, have an independent team re-code them with the same published taxonomy, and measure inter-rater agreement (for example, Cohen's kappa). If agreement falls well below conventional thresholds, the reported percentages, temporal trends, and the four MCA patterns cannot be treated as stable facts about the literature.","tokens_in":28455,"feed_emoji":"🔗","tokens_out":2693,"duration_ms":36327,"temperature":0.7,"pith_summary":"This paper claims to provide the first large-scale systematic literature review of the blockchain-and-edge-computing nexus, covering 921 papers published between 2015 and October 2022. It builds a 22-dimension, 287-attribute taxonomy and applies quantitative analysis plus multiple correspondence analysis to show how the two paradigms interact, co-evolve, and cluster into distinct research styles. A sympathetic reader should care because the result is an empirical map of an entire research area: which problems dominate, which design choices are actually used, and where the field is maturing. The paper argues that blockchain-assisted edge computing, especially for security and privacy, is the prevalent direction, while edge-assisted blockchain work remains less developed.","feed_headline":"921 papers reveal four ways blockchain meets edge computing","feed_subtitle":"A 22-dimension taxonomy and machine-learning analysis show the field splits into permissioned design, technology, proof of concept, and…","key_machinery":"The load-bearing apparatus is the taxonomy itself: 22 dimensions covering scope, application, problem, contribution, AI method, allocation, metrics, technology, TRL, open data, communication, evaluation, security, privacy, sustainability, blockchain platform, type, permission, consensus, chain, and reward, together containing 287 attributes. The taxonomy is constructed following a published taxonomy-development method and then applied to every paper. The quantitative engine is multiple correspondence analysis (MCA), which reduces the categorical coding into meta-dimensions; the paper retains four meta-dimensions and interprets them as the four study patterns. The taxonomy supplies the raw structure, and MCA supplies the grouping that turns manual labels into discrete patterns.","core_discovery":"The central claim is that the blockchain-edge computing literature, far from being an undifferentiated mass, separates into four distinguishable study patterns: private permissioned design, technology implementation, proof-of-concept research, and public permissionless design. The paper further claims that 75% of the reviewed papers take the perspective of blockchain assisting edge computing, while only 19% use edge computing to assist blockchain and 6% pursue a full synergy. It reports that permissioned blockchains dominate the field, that privacy is the fastest-rising problem area, and that the choice of permissioned versus permissionless design is the key determinant of how the two paradigms are combined. These claims rest on a manually constructed taxonomy and a machine-learning analysis of attribute co-occurrence across the 921 papers.","pith_inferences":["If the manual coding is reliable, the released open dataset could be used to train an automated classifier that tracks the blockchain-edge literature beyond October 2022, effectively extending the review without a full re-coding effort.","The paper's cutoff predates the recent surge of Decentralized Physical Infrastructure Networks; applying the same taxonomy to 2023-2025 publications would test whether the four patterns persist or whether the permissionless, incentive-heavy pattern grows.","The observed dominance of blockchain-for-edge security may partly reflect publication incentives in the security research community rather than engineering demand, meaning the 75/19/6 perspective split should not be read as a measure of real-world deployment.","The four MCA patterns could serve as sampling strata for future deep-dive meta-analyses, letting reviewers compare findings within each pattern rather than averaging across a heterogeneous corpus."],"forward_implications":["Researchers can use the taxonomy as a checklist to position new work and to notice under-explored combinations, such as the small 6% share of studies that genuinely integrate both paradigms.","The prevalence of permissioned blockchains in about 75% of papers implies that blockchain-edge systems are mostly studied in controlled, consortium-style settings, while scalability-focused work almost always assumes permissioned design.","The sharp rise in privacy-related papers, with a 74% relative increase from 2020 to 2021, signals that privacy protection is becoming a central motivation for combining blockchain with edge computing.","Edge-assisted blockchain research (Perspective 2) is less mature in both methodological contribution and technology readiness, marking it as a comparatively open research direction.","The four MCA patterns give the field a compact vocabulary: private permissioned design, technology implementation, proof-of-concept research, and public permissionless design, each with distinct technology and consensus choices."],"supporting_citations":[{"why":"Supplies the PRISMA 2020 reporting guideline that structures the systematic review protocol.","marker":"[189]"},{"why":"Provides the taxonomy development method on which the paper's 22-dimension classification is built.","marker":"[89]"},{"why":"Contributes a prior distributed-ledger taxonomy whose dimensions and attributes are adapted for the blockchain-edge domain.","marker":"[8]"},{"why":"Serves as a smaller existing systematic review of blockchain-fog integration applications that this work extends in scale and scope.","marker":"[87]"},{"why":"Provides another smaller systematic review of integration purposes, used as a comparison baseline for sample size and number of features.","marker":"[88]"},{"why":"Represents an earlier mini systematic review focused on cybersecurity in fog computing, used as a baseline for limitation of perspective.","marker":"[67]"},{"why":"Represents a prior systematic review of security issues in cloud/edge computing with blockchain, used to contrast against the broader multi-perspective approach.","marker":"[68]"}],"fun_headline_variants":["Four patterns emerge from 921 blockchain-edge papers","Blockchain meets edge: 921 papers, four patterns","Large-scale review finds 4 blockchain-edge study patterns","Machine learning on 921 papers reveals 4 blockchain-edge designs","Taxonomy of 921 papers: 4 ways to combine blockchains and edge"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The entire quantitative analysis depends on the five authors' manual coding of 921 papers into the taxonomy being accurate and consistent, and no inter-rater reliability measure is reported for that coding.","fun_headline_variants_meta":{"raw":{"variants":["Four patterns emerge from 921 blockchain-edge papers","Blockchain meets edge: 921 papers, four patterns","Large-scale review finds 4 blockchain-edge study patterns","Machine learning on 921 papers reveals 4 blockchain-edge designs","Taxonomy of 921 papers: 4 ways to combine blockchains and edge"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00019,"raw_usage":{"total_tokens":1317,"prompt_tokens":902,"completion_tokens":415,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":518,"completion_tokens_details":{"reasoning_tokens":331}},"tokens_in":518,"tokens_out":415,"duration_ms":5169,"temperature":1.0,"reasoning_tokens":331,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T05:04:24.800343+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a random sample of roughly 100 of the 921 papers, have an independent team re-code them with the same published taxonomy, and measure inter-rater agreement (for example, Cohen's kappa). If agreement falls well below conventional thresholds, the reported percentages, temporal trends, and the four MCA patterns cannot be treated as stable facts about the literature.","supporting_citations":[{"cited_title":"The prisma 2020 statement: an updated guideline for reporting systematic reviews.International journal of surgery, 88:105906, 2021","cited_arxiv_id":null,"evidence_quote":"Supplies the PRISMA 2020 reporting guideline that structures the systematic review protocol."}],"review_version":1}