Alexandria Digital Research Library

Harvesting Geospatial Intelligence from Geotagged Social Media Data : A New Type of Early Warning System against North Korea

Author:
Kim, Jeong-Hyun
Degree Grantor:
University of California, Santa Barbara. Geography
Degree Supervisor:
Keith C. Clarke
Place of Publication:
[Santa Barbara, Calif.]
Publisher:
University of California, Santa Barbara
Creation Date:
2015
Issued Date:
2015
Topics:
Geographic information science and geodesy and Geography
Keywords:
Flickr
Early warning system
Geospatial intelligence
Geotagged data
North Korea
Social media
Genres:
Online resources and Dissertations, Academic
Dissertation:
M.A.--University of California, Santa Barbara, 2015
Description:

After the Korean War in 1950, the Korean Peninsula was divided into North and South Korea. From that time on, a hostile stand-off has persisted for more than 60 years. After Kim Jong-un come to power, North Korea's provocations, such as the Yeon-Pyong Island bombardment and the attack of a Chen-an South Korean navy ship, reached a new level. North Korea is still threatening international societies by developing Weapons of Mass Destruction (WMD) such as nuclear weapons and Intercontinental Ballistic Missiles. Moreover, as we saw from the 'Sony hack incident' where Sony Pictures Entertainment was hacked by North Korea in 2014 because of the film 'The Interview,' North Korea is now concentrating on cyber warfare. Thus, because of these incidents, the need for an advanced Early Warning System (EWS) against North Korea is increasing.

This thesis demonstrates the potential of social media data as a new source to collect Geo-spatial Intelligence (GeoINT) for an advanced EWS. In this thesis, structural analysis, based on geospatial data and content analysis utilizing images and text tags were conducted using two different datasets, Flickr and Twitter. First, a new approach for geotagged image processing was designed using the K-means and Spectral clustering algorithm using 3,981 North Korea Flickr images over a period of 11 years. Principal Components Analysis (PCA) demonstrated that image entropy is an effective input parameter for image characterization. We concluded that the Spectral clustering algorithm is suitable for selecting outlier images. In addition, Centrographic statistics were applied for the analysis of spatial and temporal dynamics. As a result, geotagged social media data revealed their user's behavior and interest by time and space.

Second, 12,394 tweets from all over the world during the 'Sony hack incident' were analyzed for a case study. Study findings revealed that Twitter data provides a real time reflection of user's interest, event, place and time. Although North Koreans cannot use the Internet or social media applications, there were 11 tweets from North Korea and all of the locations for tweets were identified by Google Earth. By doing so, the potential flaws in using social media data, such as IP address hacking, were found.

For the tag analysis of Flickr and Twitter, all tags were counted and high-frequency tags were extracted and visualized with Word Cloud techniques. Lastly, various maps were generated using several visualization techniques.

This study is the first study about North Korea using geotagged social media data. It indicates that social media data has a rigorous potential and can be a new source for an EWS. This new approach method with social media data can be applied to other closed countries.

Physical Description:
1 online resource (112 pages)
Format:
Text
Collection(s):
UCSB electronic theses and dissertations
ARK:
ark:/48907/f3348hjh
ISBN:
9781339084398
Catalog System Number:
990045715810203776
Rights:
Inc.icon only.dark In Copyright
Copyright Holder:
Jeong Hyun Kim
Access: This item is restricted to on-campus access only. Please check our FAQs or contact UCSB Library staff if you need additional assistance.