2025, Vol.21, No.2

Journal of Software Forensics 2025 KCI Impact Factor : 0.3

pISSN : 3092-541X / eISSN : 3140-071X
https://journal.kci.go.kr/ksavs

Disputes over Disputes over software completeness frequently arise in development contracts and outsourced projects due to the ambiguity surrounding the implementation status of contractual requirements. This paper proposes a novel framework to address this issue by introducing an objective and automated approach to software completeness evaluation. The framework utilizes NLP techniques to extract requirements from contract documents and matches them with implemented source code using deep learning-based analysis. By quantitatively assessing the degree to which the implemented code fulfills the specified requirements, the framework reduces reliance on the subjective judgments of human evaluators. It supports both quantitative and qualitative evaluation criteria, offering improvements in assessment reliability, cost-efficiency, and time-effectiveness. A case study is presented to validate the framework’s effectiveness, demonstrate its applicability as a practical tool for appraisal, quality assurance, and development monitoring.

This paper examines the copyrightability of library files—components of computer programs—and practical issues in their application, focusing on a recent Supreme Court ruling (2024do3533). While copyright infringement is typically assessed through source and executable code, other expressive elements, including library files in binary form, may also qualify for protection if they meet originality requirements. However, binary code’s limited readability hinders the evaluation of originality, leading to its exclusion from substantial similarity analysis in practice. To address this, the paper proposes court-led confirmation procedures for appraisal targets, improved quantitative similarity methods, and clearer criteria for assessing originality in library files.

With the increasing use of OTT content, there is a growing demand for copyright protection across diverse playback environments. In particular, technologies capable of identifying original audio even under the presence of unstructured noise—such as white noise or conversational background noise—are gaining importance. This paper proposes a robust method for extracting and recognizing audio DNA that remains reliable in such noisy conditions. By leveraging feature extraction based on MEL spectrograms, the proposed system generates a dual-stage fingerprint structure consisting of coarse and fine features. Recognition performance was evaluated under various parameter combinations. Experimental results demonstrate that by optimizing FFT length and feature dimension, the proposed method achieves high accuracy. The results confirm that the approach can be effectively applied to real-time content authentication and copyright protection systems in OTT service environments.