Hardware and software system for verifiably processing bibliometric data

Abstract

The invention relates to computing. A hardware and software system for verifiably processing bibliometric data, ranking publications and generating a cumulative rating comprises a document digitization and recognition station for processing physical media, a data receiver, a graph processing server, a parallel analytics cluster for generating analytical signals, a verification device with TEE support, a hardware API gateway, a printer for printing secure attestations to provide a report in physical format, and a central data storage array, as well as a user terminal for accessing the API gateway. The invention provides not only for the increased accuracy of a rating, but also for the verifiable validity and integrity thereof.

Description

[0001] A hardware and software system for verified processing of bibliometric data, ranking of publications and generation of an integrated rating system

[0002] The invention relates to secure data processing hardware, specifically to distributed computing systems implemented as physically isolated but interconnected devices designed to generate verifiable analytical metrics from multiple sources, including both digital and physical (paper) media. Technically, this hardware system addresses the problem of ensuring the integrity and verifiability of the results of complex multifactor calculations performed on large data sets. It finds application in areas requiring not only calculation results but also hardware-guaranteed confirmation that the calculation process itself was protected from unauthorized access, as exemplified by the processing of bibliometric information to construct objective and tamper-resistant rankings.

[0003] Existing systems for analyzing scientific activity are typically implemented as software applications running on general-purpose servers. These systems rely on metrics (such as Impact Factor, SJR), which, due to their purely software architecture, have significant technical limitations. They primarily process direct, unfiltered citation indicators, making their computational processes vulnerable to data manipulation, such as through anomalous self-citations.

[0004] A system for ranking nodes in linked databases is known, using network analysis, for example, based on the PageRank algorithm. This system assigns importance ranks to nodes in a linked database, for example, in any database of documents containing citations, on the World Wide Web, or in any other hypermedia database. The document rank is calculated based on the ranks of the documents citing it. In addition, the document rank is calculated based on a constant representing the probability that a browser browsing the database will randomly navigate to the document. This method is particularly useful for improving the performance of search results in hypermedia databases, such as the World Wide Web, whose documents vary widely in quality (US No. 6285999 Bl, September 4, 2001).However, the technical implementation of such systems typically does not include mechanisms for weighting graph connections based on source credibility or for taking into account thematic relevance, treating all connections as equal. This is an architectural flaw that reduces the reliability of the final result.

[0005] The closest analogue is the Submission Decision Support System, which aims to create a unified decision support system for manuscript submissions. It includes three main subsystems: a decision factor filtering system, a submission decision support system, and a decision model validation system (US No. 20100106669 Al, April 29, 2010).

[0006] However, this system lacks a key technical component: a module for automated verification of the authenticity of the source data.

[0007] Furthermore, the most significant technical drawback of existing systems is the lack of physical means to ensure and verify the integrity of the computational process itself. Since critical calculations are performed in an unprotected environment on standard hardware, the end user receives the calculation result but cannot obtain technical, hardware-guaranteed proof that the calculation was performed by a correct, unmodified algorithm and based on verified data. Thus, existing systems function as "black boxes" at the software level, failing to address the technical problem of creating a provably reliable and tamper-resistant analytical indicator whose integrity can be physically verified.

[0008] The objective of the invention is to create a distributed hardware and software system that overcomes the limitations of existing software analogues by solving the following technical subproblems at the hardware level:

[0009] - creation of a multi-component hardware data processing pipeline by constructing a physical infrastructure that includes specialized devices for each processing stage, which ensures hardware separation and optimization of computing tasks;

[0010] - Integration of specialized hardware filters into the architecture to improve noise immunity and reliability of the computing process through the use of a dedicated data quality control node optimized for automated detection and correction of anomalies in the input data stream; - Provable computational integrity at the hardware level to eliminate the "black box" problem inherent in software systems, with the goal of performing the final, most critical stage of the calculation within a verification device using a hardware-protected environment (TEE). This transfers integrity guarantees from the software level to the physical, hardware level of the processor;

[0011] - Implementation of physical and cryptographic verification of the result to create a mechanism that provides not only the final result via a hardware API gateway but also cryptographic confirmation (certificate) that the calculation was performed correctly in a secure hardware environment. This enables external verification and, when using a printer, will provide tangible proof of the calculation's integrity.

[0012] The technical result is the creation of a system that ensures not only increased rating accuracy, but also its provable reliability and integrity, thanks to an architecture that implements multi-factor analysis with automatic anomaly filtering and critical calculations performed in a hardware-protected environment, ensuring transparency, the possibility of external verification, and eliminating the risk of manipulation.

[0013] This technical result is achieved by proposing a hardware and software system (HSS) for the verified processing of bibliometric data, publication ranking, and integrated rating generation. The system is built on the interaction of specialized physical devices that support the full processing cycle: from digitizing paper media to providing a cryptographically verified final rating. Thus, the proposed HSS includes:

[0014] - a station for digitalization and recognition of documents for processing physical media (for example, paper journals), consisting of a high-speed document scanner, a dedicated processor-computing unit and a verification terminal;

[0015] - a data receiving device, which is a dedicated network server that receives data from a document digitization and recognition station and external sources to transmit a unified data stream further along the conveyor;

[0016] - a graph processing server equipped with a large amount of RAM and fast storage devices, converting data streams into a weighted citation graph structure; - a parallel analytics cluster for generating analytical signals, consisting of several physical server nodes, processing data in parallel to generate three independent indicators:

[0017] • a network analysis node for calculating the Citation Influence Score (CIS) through iterative analysis of the graph structure;

[0018] • content analysis node for analyzing publication texts and calculating the thematic similarity score (Quality-adjusted Similarity Score, QSS);

[0019] • a data quality control node that functions as a hardware filter, analyzing the graph for atypical citation patterns and generating a Data Integrity Score (DIS);

[0020] - a verification device with TEE support, implemented as a physical server whose processor supports a hardware trusted execution environment (TEE);

[0021] - hardware API (Application Programming Interface) gateway, which is a device that manages all external requests to provide a verified result and publish the final QJR rating;

[0022] - a printer for printing secure certificates for outputting a report onto a tangible medium;

[0023] - a central storage array, which is a network storage system (Storage Area Network, SAN or Network Attached Storage, NAS) with RAID (Redundant Array of Independent Disks) support for fault tolerance, and

[0024] - a user terminal that accesses the API gateway.

[0025] Description of the invention with reference to the figure:

[0026] Fig. 1 shows a block diagram of a PAC with verified computational integrity.

[0027] Embodiments of the invention will now be described in detail with reference to the accompanying figure.

[0028] Thus, according to Fig. 1, the proposed hardware and software complex (HSC) for verified processing of bibliometric data, ranking of publications and generation of an integral rating includes:

[0029] - a document digitization and recognition station (software) for processing physical media (for example, paper journals), consisting of a high-speed document scanner, a dedicated processor-computing unit and a verification terminal;

[0030] - a data receiving device (120), which is a dedicated network server that receives data from a document digitization and recognition station and external sources for transmitting a unified data stream further along the conveyor;

[0031] - graph processing server (130), equipped with a large amount of RAM and fast storage devices, transforming data streams into a weighted citation graph structure;

[0032] - a parallel analytics cluster (140) for generating analytical signals, consisting of several physical server nodes that process data in parallel to generate three independent indicators:

[0033] • network analysis node (141) for calculating the Citation Influence Score (CIS) through iterative analysis of the graph structure;

[0034] • content analysis node (142) for analyzing publication texts and calculating the thematic similarity score (Quality-adjusted Similarity Score, QSS);

[0035] • a data quality control node (143), which functions as a hardware filter, analyzing the graph for the presence of atypical citation patterns and generating a data integrity score (DIS);

[0036] - a verification device with TEE support (150), implemented as a physical server whose processor supports a hardware trusted execution environment (TEE);

[0037] - hardware API gateway (170), which is a device that manages all external requests to provide a verified result and publish the final QJR (Quality Journal Rank);

[0038] - printer for printing secure certificates (180) for outputting a report onto a tangible medium;

[0039] - a central storage array (160), which is a network storage system (SAN or NAS) with RAID support for fault tolerance, and

[0040] - user terminal (190) that accesses the API gateway (170).

[0041] The proposed PAC operates and is used as follows:

[0042] The technical result is achieved through the interaction of physical devices. Each component of the hardware and software system is a specialized device configured to perform a specific step in the verified data processing pipeline.

[0043] Architecture and operation of hardware components:

[0044] 1. Data input and structuring devices:

[0045] Processing begins with devices designed to collect and initially process information from physical and digital sources.

[0046] Document Digitization and Recognition Station (110): This device is a hardware workstation for data input from paper media. Physically, the station consists of:

[0047] • Industrial Document Scanner: High-speed document feeder capable of handling high volumes of printed material.

[0048] • Dedicated Processing Unit: A compact computer paired with the scanner that runs software for Optical Character Recognition (OCR) and document structure analysis. This unit is hardware optimized for image processing tasks.

[0049] • Verification Terminal: A touch screen monitor with a keyboard that allows the operator to control the recognition process and make manual adjustments to the extracted metadata (title, authors, journal, etc.) before sending the data to the system.

[0050] Data reception device (120): This is a specialized network gateway (appliance) designed as a rack-mounted server. Its technical purpose is data aggregation. The device is equipped with multiple high-speed network interface cards (NICs) for simultaneous data reception from the digitization station (110) and from external digital APIs (e.g., Web of Science, Scopus) via secure protocols (Transport Layer Security, TLS). A hardware firewall built into the device provides perimeter protection.

[0051] Graph Processing Server (130): Data from the device (120) is fed to this server. It is configured for intensive operations with graph databases (e.g., Neo4j). Its hardware configuration includes a multi-core processor, a significant amount of RAM (256 GB or more), and an array of solid-state drives (NVMe SSDs) operating in RAID mode to ensure the maximum read / write speed necessary for building and updating the weighted citation graph in real time. The parallel analytics cluster (140) is used for efficient multivariate analysis; the citation graph is processed in a cluster, which is a single chassis with multiple independent server blades. The cluster nodes are interconnected by a high-speed internal bus (e.g., InfiniBand) to minimize data exchange latency. Each node is configured to perform a specific task:

[0052] • Network Analysis Node (141): Optimized for iterative computation, it analyzes graph structure to calculate the Citation Impact Score (CIS).

[0053] • Content Analysis Node (142): Equipped with a Graphics Processing Unit (GPU) to accelerate computations related to Natural Language Processing (NLP) models and calculates the topic similarity score (QSS).

[0054] • Data Quality Node (143): Uses a processor with an extended vector instruction set to efficiently apply machine learning algorithms to detect anomalous citation patterns and calculate a Data Integrity Score (DIS).

[0055] Verification Device and Secure Computing (150): This is the core of the entire system, ensuring the provable integrity of computations.

[0056] • Technical basis: the device (150) is a physically secure server whose central processor (e.g., Intel® Xeop® with SGX technology) supports the creation of a trusted execution environment (TEE). This environment creates a secure enclave - an isolated area on the processor chip itself. Code and data inside the enclave are encrypted and inaccessible for inspection or modification by any other software, including the host operating system.

[0057] • Secure Operations: Two critical operations are performed within this hardware enclave: 1) the calculation of the source's base Journal Credibility Index (JCI); 2) the final aggregation of the CIS, QSS, and DIS scores into the final QJR. The process is iterative. The server (130) may use previously calculated JCI values ​​to initially construct the graph. During each processing cycle, the verification appliance (150) calculates updated, verified JCI values, which are passed back to the Graph Processing Server (130) to refine the graph weights in subsequent calculation cycles. When the system is started for the very first time, no "previously calculated" values ​​exist. In this case, the Graph Processing Server (130) uses the initial (starting) JCI value. Typically, this is a neutral value (e.g., JCI = 1.0) that is assigned to all sources by default.

[0058] • Hardware authentication: To authorize changes to the code running inside the enclave (for example, when updating the rating calculation formula), the device (150) is equipped with a USB port for connecting a physical authentication key. Administration operations can only be performed after presenting such a key.

[0059] • Attestation: The device (150) is capable of generating a cryptographic report (attestation) signed with the processor's unique hardware key. This report allows any external party to cryptographically verify that a computation was performed by genuine, unmodified code inside the genuine hardware enclave.

[0060] 2. Storage, output and interaction devices:

[0061] The central storage array (160) is a network-attached storage system (SAN or NAS) with RAID support for fault tolerance. It stores the final QJR ratings and other service information.

[0062] The hardware API gateway (170) is a physical device that manages all external requests to the system. It performs hardware encryption acceleration (TLS offloading) and routes requests to the storage array (160) for data retrieval. The gateway also provides access to cryptographic certificates generated by the device (150).

[0063] Secure Certificate Printer (180): A specialized printing device connected to the gateway (170). It is designed to print cryptographic certificates onto a physical medium, creating a tangible, tamper-proof proof of the integrity of the computations.

[0064] The user terminal (190) is any end user device (personal computer, workstation) that, through a standard web browser, accesses the web interface provided by the API gateway (170) to visualize the ratings and their components.

[0065] The proposed software system operates as follows: Data processing begins with physical and digital sources. Paper media are sent to the document digitization and recognition station (DRS), which converts them into digital format. This stream, along with data from external digital sources (e.g., Web of Science, Scopus), is sent to the data receiving device (120).

[0066] The unified data stream is transmitted to the graph processing server (130), which generates a weighted citation graph. The graph then serves as the basis for parallel processing in the parallel analytics cluster (140). The cluster consists of three independent hardware nodes: a network analysis node (141), which calculates the citation impact score (CIS); a content analysis node (142), which calculates the subject similarity score (QSS); and a data quality control node (143), which generates the data integrity score (DIS).

[0067] These three metrics (CIS, QSS, DIS) are passed as input to the system's central node—a TEE-enabled verification device (150). It is within this device (150), in an isolated and secure hardware environment, that the final aggregation and calculation of the final rank (QJR) occurs.

[0068] The calculated rank (QJR) is stored in the central storage array (160). Simultaneously, the device (150) generates a cryptographic certificate confirming the integrity of the calculations. This certificate can be transferred to the secure certificate printer (180) to create a physical copy.

[0069] Access to the final ratings from the repository (160) and to the verification certificate is provided through the hardware API gateway (170). End user interaction with the system is accomplished through the user terminal (190), which accesses the API gateway (170).

[0070] Thus, the proposed software system implements a fully functional data processing and verification cycle—from digitization and aggregation of information to obtaining the final rating and its confirmation in a secure environment. The system's architecture ensures a high degree of reliability, transparency, and provable validity of results, making it applicable to tasks where the objectivity and verifiability of calculations are critical.

Claims

CLAUSES OF THE INVENTION A hardware and software system for the verified processing of bibliometric data, ranking of publications, and generation of an integrated rating, which is built on the interaction of devices that provide a full processing cycle from the digitization of paper media to the provision of a cryptographically verified final rating, characterized in that the hardware and software system additionally includes? - a station for digitalization and recognition of documents for processing physical media, consisting of a high-speed document scanner, a dedicated processor-computing unit and a verification terminal; - a data receiving device, which is a dedicated network server that receives data from a document digitization and recognition station and external sources to transmit a unified data stream further along the conveyor; - a graph processing server equipped with a large amount of RAM and fast storage devices, which transforms data streams into a weighted citation graph structure; - a parallel analytics cluster for generating analytical signals, consisting of physical server nodes that process data in parallel to generate three independent indicators: • a network analysis node for calculating the Citation Influence Score (CIS) through iterative analysis of the graph structure; • content analysis node for analyzing publication texts and calculating the thematic similarity score (Quality-adjusted Similarity Score, QSS); • a data quality control node that functions as a hardware filter, analyzing the graph for atypical citation patterns and generating a Data Integrity Score (DIS); - a verification device with TEE support, implemented as a physical server whose processor supports a hardware trusted execution environment (TEE); - a hardware API gateway, which is a device that manages all external requests to provide a verifiable result and publish the final QJR rating; - a printer for printing secure certificates for outputting a report onto a tangible medium; - a central storage array, which is a network-attached storage system (SAN or NAS) with RAID support for fault tolerance, and - a user terminal that accesses the API gateway.

Images