Data Source: OpenAlex for Cybersecurity Technology Convergence Study

Table of Links

Abstract and 1. Introduction

2 Related Work and 2.1 Technology Convergence Approaches

2.2 Technology Convergence Measurements

2.3 Technology Convergence Models

3 Data

4 Method and 4.1 Proximity Indices

4.2 Interpolation and Fitting Data

4.3 Clustering

4.4 Forecasting

5 Results and Discussion and 5.1 Overall Results

5.2 Case Study

5.3 Limitations and Future Works

6 Conclusion and References

Appendix

3 Data

This study utilizes data from OpenAlex, a comprehensive repository of research papers, authors, and institutions, which was created by drawing aspirations of the ancient Library of Alexandria [29]. OpenAlex employs an automated system that assigns a set of Wikidata concepts to each paper from a pool of 65,000 concepts. This is accomplished through a multi-class deep learning classifier trained on the Microsoft Academic Graph. Each concept is assigned a score between 0 and 1, indicating the paper’s relevance to that particular concept [29].

Data Extraction and Processing

Data relevant to the evolution of encryption technologies was extracted and processed into time series for analysis and forecasting. The focus was on scientific concepts related to 36 encryption technologies identified by experts through the Delphi method [30]. Among the 36 technologies, only 25 were found in OpenAlex, forming the basis of our analysis. Papers not relevant to these 25 technologies were assigned a zero score. It was observed that over 90% of the papers had a zero attribution score, affirming the importance of the scoring system (refer to Figure 5 in the appendix).

Data Refinement

Due to the incomplete data in OpenAlex, papers without references, constituting half of the dataset, were removed to prevent bias. The distribution of these papers across technologies was uniform, ruling out any potential data skew (see Figure 6 in the appendix). Additionally, we excluded another 5% of the remaining papers that were not linked to any concepts. To address anomalies like the overrepresentation of papers published in January, we corrected them by evenly redistributing throughout the year. Duplicate entries were resolved by keeping the version with the most comprehensive information.

Data Enhancement

Keywords were assigned to each paper using KeyBert [1], a model that identifies keywords based on their similarity to the text, utilizing ’cosine similarity’ to indicate keyword relevance. Although a substantial portion of keywords was unique, common keywords played a crucial role in our analysis (see Figure 2 in the appendix).

Author Influence

We calculated the h-index for authors in the encryption field from 2002 to 2022, generating both incremental and non-incremental indices on both a monthly and yearly basis. The incremental indices, offering a cumulative perspective of an author’s impact, were considered a more accurate representation of an author’s significance in the field (refer to Figures 3 and 4 in the appendix).

Data Consolidation

The completed dataset included different elements like h-indices, references, and keywords. These were organized into a detailed data frame with columns such as ’paper,’ ’keyword,’ ’cosine similarity,’ ’title,’ ’publication date,’ ’abstract,’ ’year,’ ’month,’ ’author,’ ’referenced works,’ ’concepts,’ ’score concepts,’ ’yearly H index not incremental,’ ’yearly H index incremental,’ ’monthly H index incremental,’ and ’monthly H index not incremental.’

[1] https://maartengr.github.io/KeyBERT/api/keybert.html