[Paper Review] A New High-Performance Approach to Approximate Pattern-Matching for Plagiarism Detection in Blockchain-Based Non-Fungible Tokens (NFTs)

Ciprian Pungilă, Darius Galiş+1 more|arXiv (Cornell University)|May 28, 2022

Academic integrity and plagiarism4 citations

TL;DR

This paper proposes a novel, high-performance approximate pattern-matching approach for plagiarism detection in blockchain-based NFTs using a non-deterministic finite automaton (NDFA) with sliding windows and local thresholds. It achieves up to 9.5× speed-up over classical similarity measures like Levenshtein, while maintaining high accuracy, making it suitable for real-time NFT intellectual property protection.

ABSTRACT

We are presenting a fast and innovative approach to performing approximate pattern-matching for plagiarism detection, using an NDFA-based approach that significantly enhances performance compared to other existing similarity measures. We outline the advantages of our approach in the context of blockchain-based non-fungible tokens (NFTs). We present, formalize, discuss and test our proposed approach in several real-world scenarios and with different similarity measures commonly used in plagiarism detection, and observe significant throughput enhancements throughout the entire spectrum of tests, with little to no compromises on the accuracy of the detection process overall. We conclude that our approach is suitable and adequate to perform approximate pattern-matching for plagiarism detection, and outline research directions for future improvements.

Motivation & Objective

• To address the growing challenge of plagiarism in digital assets, especially within blockchain-based NFT ecosystems.
• To improve the performance of approximate pattern-matching for plagiarism detection without sacrificing accuracy.
• To enable real-time, scalable detection of near-duplicate content in large-scale NFT and digital content platforms.
• To reduce false positives and false negatives in similarity-based plagiarism detection using adaptive thresholding.
• To design a scalable, high-throughput solution suitable for integration with NFT-enabled blockchain systems.

Proposed method

• The approach employs a non-deterministic finite automaton (NDFA) to index and match patterns efficiently across input data.
• A sliding window mechanism is applied to process input data in chunks, enabling local similarity scoring at each node level.
• Local thresholds are set at each automaton node to detect partial matches early, reducing unnecessary full-sequence comparisons.
• The method integrates multiple similarity measures—Euclidean, Hamming, and Levenshtein—within the NDFA framework for flexible pattern matching.
• The automaton stores all suffixes at each node, enabling efficient detection of approximate matches through dynamic programming-like state transitions.
• Performance is optimized by processing patterns in parallel and leveraging the NDFA’s ability to handle multiple potential matches simultaneously.

Experimental results

Research questions

• RQ1 How can approximate pattern-matching be accelerated for plagiarism detection in large-scale NFT and digital content systems?
• RQ2 What is the trade-off between speed and accuracy when using local thresholds and sliding windows in similarity-based pattern matching?
• RQ3 To what extent can an NDFA-based automaton outperform classical similarity measures like Levenshtein, Hamming, and Euclidean distance in throughput and precision?
• RQ4 How do varying pattern lengths and window sizes affect the performance and accuracy of the proposed method?
• RQ5 Can the proposed approach be scaled for real-time deployment in blockchain-based NFT platforms with minimal storage overhead?

Key findings

• The proposed NDFA-based approach achieved up to 9.5× higher throughput for Levenshtein similarity compared to classical methods, especially with longer patterns.
• Euclidean similarity maintained a constant 1 false negative across all threshold values, indicating stable detection reliability.
• Hamming and Levenshtein similarities achieved zero false positives when the threshold was set to 0.85 or higher, indicating high precision at high thresholds.
• The method showed a 9× speed-up for Euclidean similarity and a 3× speed-up for Hamming distance, demonstrating consistent performance gains.
• The approach achieved significant speed-up even with longer patterns, indicating strong scalability with input size.
• Despite high throughput, the method incurs higher memory usage due to full suffix storage at each node, prompting future work on storage optimization.