arXiv subjects

Katsunari Yoshioka

Publications and source records attributed to Katsunari Yoshioka.

3 recordsLinked to original sources

arXiv2026

Understanding Interpretation Difficulty in Harmful Online Communication: Insights from Cybercrime Communities

Harmful online communication often contains slang, coded terms, abbreviations, and community-specific expressions, which make messages difficult to interpret. This paper presents an exploratory study of interpretation difficulty in Discord chats related to cybercrime. We construct reference interpretations of purposefully selected difficult messages, which were reviewed by an expert. We then use them to evaluate human and large language model (LLM) interpretations under different context conditions. The results show that local context alone is often insufficient for humans, while external knowledge and extended conversational context substantially improve human interpretation. For LLMs, local context also improves interpretation, and the larger model performs better. We further conduct a qualitative error analysis and propose a preliminary classification of factors that make harmful chats difficult to interpret. These findings suggest that harmful-content analysis should treat interpretation as an evidence-integration problem, rather than as message-level classification alone.

cs.CL↗

arXiv2025

Am I Infected? Lessons from Operating a Large-Scale IoT Security Diagnostic Service

There is an expectation that users of home IoT devices will be able to secure those devices, but they may lack information about what they need to do. In February 2022, we launched a web service that scans users' IoT devices to determine how secure they are. The service aims to diagnose and remediate vulnerabilities and malware infections of IoT devices of Japanese users. This paper reports on findings from operating this service drawn from three studies: (1) the engagement of 114,747 users between February, 2022 - May, 2024; (2) a large-scale evaluation survey among service users (n=4,103), and; (3) an investigation and targeted survey (n=90) around the remediation actions of users of non-secure devices. During the operation, we notified 417 (0.36%) users that one or more of their devices were detected as vulnerable, and 171 (0.15%) users that one of their devices was infected with malware. The service found no issues for 99% of users. Still, 96% of all users evaluated the service positively, most often for it providing reassurance, being free of charge, and short diagnosis time. Of the 171 users with malware infections, 67 returned to the service later for a new check, with 59 showing improvement. Of the 417 users with vulnerable devices, 151 users revisited and re-diagnosed, where 75 showed improvement. We report on lessons learned, including a consideration of the capabilities that non-expert users will assume of a security scan.

cs.CR↗

arXiv2023

Canary in Twitter Mine: Collecting Phishing Reports from Experts and Non-experts

The rise in phishing attacks via e-mail and short message service (SMS) has not slowed down at all. The first thing we need to do to combat the ever-increasing number of phishing attacks is to collect and characterize more phishing cases that reach end users. Without understanding these characteristics, anti-phishing countermeasures cannot evolve. In this study, we propose an approach using Twitter as a new observation point to immediately collect and characterize phishing cases via e-mail and SMS that evade countermeasures and reach users. Specifically, we propose CrowdCanary, a system capable of structurally and accurately extracting phishing information (e.g., URLs and domains) from tweets about phishing by users who have actually discovered or encountered it. In our three months of live operation, CrowdCanary identified 35,432 phishing URLs out of 38,935 phishing reports, 31,960 (90.2%) of these phishing URLs were later detected by the anti-virus engine. We analyzed users who shared phishing threats by categorizing them into two groups: experts and non-experts. As a results, we discovered that CrowdCanary extracts non-expert report-specific information, like company brand name in tweets, phishing attack details from tweet images, and pre-redirect landing page information.

cs.CR↗