Search-Query-SoK

Search log, screening records and classification data for the systematic review behind our SoK on zero-knowledge (ZK) blockchain interoperability protocols

About

This repository documents how the literature for the SoK was found, filtered and classified. It records:

• every database query and the number of records each search returned;
• the decisions made while screening candidate papers and deployed systems, with the reason for each exclusion;
• the BIF-ZK classification of every included solution, with the reasoning behind its inclusion;
• the rules used to make those decisions.

The aim is transparency and reproducibility. Anyone should be able to re-run the searches, see why a paper or system was kept or dropped, and check each classification against the stated rules.

The review covers two kinds of sources:

• Academic literature, collected from six bibliographic databases in two search rounds.
• Deployed systems, i.e. ZK-based interoperability protocols with a practical implementation, collected through predefined web searches and an AI search agent.

Repository structure

Search-Query-SoK/
├── README.md
├── search/
│   └── search_query.md            ← queries, search links and record counts
├── screening/
│   ├── academic-candidates.csv    ← every academic record assessed in full text, with its decision
│   ├── academic-exclusions.csv    ← the excluded academic records, with reasons
│   ├── deployed-candidates.csv    ← every deployed-system candidate, with its decision
│   └── deployed-exclusions.csv    ← the excluded deployed candidates, with reasons
├── data/
│   ├── bif-zk-academic.csv        ← BIF-ZK classification of the 64 academic proposals (Table 3)
│   └── bif-zk-deployed.csv        ← BIF-ZK classification of the 27 deployed systems (Table 6)
└── codebook/
    └── classification-rules.md    ← eligibility and classification rules

search/: how the records were found

search/search_query.md is the full search log. It contains:

• the Boolean query used in all databases and how it is built (interoperability terms and ZK terms, or phrases that combine both);
• links that reopen each search with the original query and filters for ScienceDirect, Scopus, IEEE Xplore, ACM Digital Library, Web of Science and Google Scholar;
• record counts per database and round, including the five query partitions ScienceDirect needed;
• the Google Scholar sampling strategy (the first 400 results in each of five publication-year intervals);
• the deployed-system search: the 15 predefined web queries and the prompts given to the AI agent.

screening/: what was kept and what was dropped

Each *-candidates.csv file lists every record that reached the final assessment stage, together with its decision. The matching *-exclusions.csv file repeats the excluded records with the reason. The id column links the two.

File	Rows	Contents
academic-candidates.csv	342	239 round-1 and 103 round-2 records, with bibliographic data and the decision on each
academic-exclusions.csv	237	the academic records not included, with the reason
deployed-candidates.csv	70	every deployed-system candidate, how it was found and the decision on it
deployed-exclusions.csv	43	the deployed candidates not included, with the reason

data/: the final corpus

Both files start with the same columns: entry (the table row, matching table_entry in screening/), references (citation numbers in the SoK), and title and link of the primary source. They then share the classification columns, in the same order. The academic file adds authors, year, venue and round. The deployed file adds website, deployment_status and subgroup.

File	Rows	Corresponds to
bif-zk-academic.csv	64	Table 3: non-deployed academic proposals
bif-zk-deployed.csv	27	Table 6: deployed systems, with deployment status and website

Scores use full, partial and none for the filled, half-filled and empty circles in the SoK's tables, and N/A where a value could not be determined.

codebook/: how solutions were classified

codebook/classification-rules.md sets out the decision sequence:

1. eligibility tests and exclusion codes;
2. corpus assignment (deployed or non-deployed);
3. a step-by-step procedure, with tie-breakers, for the 7 BIF mechanism categories;
4. the trust-model annotation;
5. the 7 functional roles;
6. the proof objectives;
7. the proof-system profiles;
8. the open-source and security-assurance labels.

Methodology in brief

flowchart LR
    A[Academic databases<br/>6 sources, 2 rounds] --> C[Deduplication]
    C --> D[Title and abstract screening]
    D --> E[Full-text assessment<br/>screening/academic-*.csv]
    B[Deployed systems<br/>web search + AI agent] --> F[Manual assessment<br/>screening/deployed-*.csv]
    E --> G[BIF-ZK classification<br/>codebook/]
    F --> G
    G --> H[Final corpus<br/>data/]

Loading

Reproducing the searches

The links in the search log open each database with the exact query and filters that were used. Note that:

• Scopus, Web of Science and ScienceDirect require institutional access.
• Database contents change over time, so re-running a search today will likely return more records than reported here.
• Google Scholar must be repeated once per publication-year interval (2020–2021, 2022–2023, 2024, 2025, 2026), screening the first 400 results of each.