Search

With the recent proliferation of social media, the spread of disinformation and misinformation has become a global risk. This study analyzes and characterizes the relationship between human emotions and other elements (social bots and echo chambers), which are major factors in the spread of information, including disinformation and misinformation. Specifically, we obtain from X (formerly Twitter) the case that was fact-checked as disinformation in Japan and analyze the relationship between each factor in a time sequence using existing tools and methods. After confirming the similar results to previous studies that emotional posts were more likely to be spread, our analysis showed the possibility that social bots were highly influential in amplifying negative sentiment in the early stages of spread. Regarding emotions and echo chambers, we observed that the increase in negative emotions and the formation of echo chambers followed a similar time sequence. In addition, we confirmed that the amplification of negative emotions precedes the formation of echo chambers. The results of this analysis suggested the importance of detecting social bot activity in the early stages of information diffusion and controlling the amplification of negative emotions.

The proliferation of disinformation has become an issue in recent years due to the widespread use of social media. This study analyzes the recent activities of social bots in Japan, which may be used for the purpose of spreading disinformation, and clarifies the characteristics of influential social bots. Specifically, we collected data from X (formerly Twitter) on several news items that had a great impact in Japan, and analyzed the spread of information by social bots using a combination of existing tools and methods. Our analysis compared the number and percentage of social bots on X in Japanese cases to existing research analyzing those during the 2016 US presidential election, and clarified what kind of social bots are influencing the information diffusion. In all cases we examined, our analysis showed that social bot activity in Japan was more active than during the 2016 US presidential election. We also found that humans are spreading posts created by social bots, as was the case during the 2016 US presidential election. Furthermore, we confirmed that the characteristics of social bots reposted by humans on X in Japan are similar to human accounts, and it is difficult to detect them using only the profile information on the X account page.

Security analysis of a web system is complicated, and thus analysis using formal methods to describe system specification mathematically has attracted attention. Some previous studies have adopted formal methods, but their models cannot express parallel communication completely. This limitation gives rise to problems where web functions, such as a cache that stores contents, cannot be defined and attacks that forge contents cannot be analyzed. These problems are present in the Alloy-based implementations of current models that do not have the ability to express temporal logic. Therefore, we design implementation and evaluation of temporal logic in Alloy to express time series and parallel computation for web security analysis. In doing so, state transitions in the web can be expressed by fitting them in our proposed syntax. As concrete applications, we describe a web security model that includes caches and show that our proposed syntax can analyze state-of-the-art attacks, such as unauthorized access to users' account pages via caches. The source code of our proposed model in Alloy is publicly available.