KJCKorea
Journal Central

Against the backdrop of shifts in the digital economy and data-sovereignty regulation, Personal Data Stores (PDS) have emerged as a viable alternative; however, limitations persist—most notably security vulnerabilities and a lack of empirical validation in operational settings. This work identifies three principal attack surfaces—(1) authentication and access control, (2) third-party applications, and (3) leakage and inference via repetitive/compound queries—and argues for systematic, quantitative evaluation aligned with market and policy developments. We propose a security architecture that counters each vector through composite proof–based access control, static analysis with permission re-confirmation, and context-aware query monitoring with response-precision control. Scenario-based analyses indicate effectiveness in preventing impersonation and privilege-escalation attempts, preempting malicious code ingress, and detecting query-driven inference attacks at an early stage. The approach strengthens data integrity, privacy, and user agency while aligning with domestic regulatory requirements such as data portability and purpose/Scope specification. The architecture is applicable across healthcare, education, and media, and is deployable alongside standardized APIs, and authentication frameworks to support real-world adoption.

The graph edit distance algorithm is widely used for structural similarity detection due to its flexibility and minimal information loss, but its computational inefficiency on large, complex graphs—stemming from its NP-Hard nature—has led to the development of various heuristic, including the Random Walk approach, which, despite its advantages, suffers from the Tottering phenomenon that degrades efficiency; although prior methods attempt to address this, they fail to eliminate Tottering entirely, prompting this paper's proposal of a novel Random Walk technique that proactively detects and prevents Tottering before it occurs. This enables more efficient similarity checks when applied to graph structure based source code similarity detection.

The webtoon industry is undergoing innovation in creation methods with the advancement of AI. Generative AIis being utilized throughout the webtoon production pipeline – from scenario writing and drawing assistance to style transfer and multi-language translation – thereby improving creative efficiency. At the same time, as AI models learn artists’ distinctive styleswithout permission and generate content resembling existing works, copyrightcontroversies have intensified. This study examines the current state of generative AI technologies supporting webtoon creation and explores corresponding copyright protection measures. In particular, it investigates the use of AI in an on-deviceenvironment, whereby creators can safely use AI tools without their data leaving their device. Through this, we suggest technical and policy directions that enable generative AI to be used as a creative assistant in webtoons while safeguarding creator rights.