國立高雄大學圖資館 |

Language: English

Back

Leveraging Social Media to Discover ...

Delaware State University.

Leveraging Social Media to Discover Threatening Tweets Using Clustering and Association Rule Mining.

Record Type:	Electronic resources : Monograph/item
Title/Author:	Leveraging Social Media to Discover Threatening Tweets Using Clustering and Association Rule Mining.
Author:	Konudula, Lakshmi Prasanna.
Published:	Ann Arbor : ProQuest Dissertations & Theses, 2020
Description:	68 p.
Notes:	Source: Masters Abstracts International, Volume: 82-02.
Notes:	Advisor: Smolinski, Tomasz G.
Contained By:	Masters Abstracts International82-02.
Subject:	Computer science.
Online resource:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=27957871
ISBN:	9798662486027

Leveraging Social Media to Discover Threatening Tweets Using Clustering and Association Rule Mining.
Konudula, Lakshmi Prasanna.

Leveraging Social Media to Discover Threatening Tweets Using Clustering and Association Rule Mining. - Ann Arbor : ProQuest Dissertations & Theses, 2020 - 68 p.

Source: Masters Abstracts International, Volume: 82-02.

Thesis (M.S.)--Delaware State University, 2020.

This item must not be sold to any third party vendors.

Online social media have become a popular communication platform through which users can express their opinions by posting short messages and connect with millions of other users instantly. Twitter is one of the most popular of such online platforms. On Twitter, a user can post a 140-character-short message about politics, reforms, events, marketing, and a million more things. Sadly, Twitter has become a prominent platform to start cyber attacks, spam feeds, introduction of malware, and much more. An innocent user tends to follow another Twitter account, open shared URLs, or use trending hashtags in the account's tweets, which may ultimately lead the user to suffer from cyber attacks. On the other hand, Twitter has become an excellent platform for researchers to extract threat intelligence information in real-time.There are several supervised learning techniques to detect threats from social media platforms. However, obtaining a labeled data set to train a supervised model is very expensive and requires human intervention. Protecting organizations from these attacks need up-to-date knowledge about the changing lingo in social media, which is necessary to detect these human or cyber threats. Constantly labeling data with the changing language is tedious, and often simply infeasible.This thesis presents a framework based on unsupervised learning for modeling of all types of threats on Twitter. It introduces an automated method to detect human threats, cyber threats, and general unhappiness of Twitter users. The proposed framework transforms tweets into vectors using the word2vec neural network, and groups similar tweets into clusters by the means of hierarchical clustering. Then, association rule mining and word clouds are used to determine the topic within each cluster. Finally, sentiment analysis is performed to estimate the general polarity of the tweets within each cluster: positive, negative, or neutral.The proposed approach was tested on over 1 million tweets spanning a period of 36 months, scraped from twitter using public API keys. The study explored various settings and parameters of the Agglomerative Hierarchical Clustering (AHC) algorithm, as well as all of the other techniques employed in the pipeline, and showed that it was possible to determine the optimal cutoff levels on the dendrogram, which produced clusters capable of grouping together tweets associated with hate speech, cyber threats, or human threats, as well as benign topics.

ISBN: 9798662486027Subjects--Topical Terms:

199325
Computer science.
Subjects--Index Terms:

Cyber threat

Leveraging Social Media to Discover Threatening Tweets Using Clustering and Association Rule Mining.
LDR:03690nmm a2200385 4500 001 594557
005 20210521101653.5
008 210917s2020 ||||||||||||||||| ||eng d
020 $a 9798662486027
035 $a (MiAaPQ)AAI27957871
035 $a AAI27957871
040 $a MiAaPQ $c MiAaPQ
100 1 $a Konudula, Lakshmi Prasanna. $3 886565
245 1 0 $a Leveraging Social Media to Discover Threatening Tweets Using Clustering and Association Rule Mining.
260 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2020
300 $a 68 p.
500 $a Source: Masters Abstracts International, Volume: 82-02.
500 $a Advisor: Smolinski, Tomasz G.
502 $a Thesis (M.S.)--Delaware State University, 2020.
506 $a This item must not be sold to any third party vendors.
520 $a Online social media have become a popular communication platform through which users can express their opinions by posting short messages and connect with millions of other users instantly. Twitter is one of the most popular of such online platforms. On Twitter, a user can post a 140-character-short message about politics, reforms, events, marketing, and a million more things. Sadly, Twitter has become a prominent platform to start cyber attacks, spam feeds, introduction of malware, and much more. An innocent user tends to follow another Twitter account, open shared URLs, or use trending hashtags in the account's tweets, which may ultimately lead the user to suffer from cyber attacks. On the other hand, Twitter has become an excellent platform for researchers to extract threat intelligence information in real-time.There are several supervised learning techniques to detect threats from social media platforms. However, obtaining a labeled data set to train a supervised model is very expensive and requires human intervention. Protecting organizations from these attacks need up-to-date knowledge about the changing lingo in social media, which is necessary to detect these human or cyber threats. Constantly labeling data with the changing language is tedious, and often simply infeasible.This thesis presents a framework based on unsupervised learning for modeling of all types of threats on Twitter. It introduces an automated method to detect human threats, cyber threats, and general unhappiness of Twitter users. The proposed framework transforms tweets into vectors using the word2vec neural network, and groups similar tweets into clusters by the means of hierarchical clustering. Then, association rule mining and word clouds are used to determine the topic within each cluster. Finally, sentiment analysis is performed to estimate the general polarity of the tweets within each cluster: positive, negative, or neutral.The proposed approach was tested on over 1 million tweets spanning a period of 36 months, scraped from twitter using public API keys. The study explored various settings and parameters of the Agglomerative Hierarchical Clustering (AHC) algorithm, as well as all of the other techniques employed in the pipeline, and showed that it was possible to determine the optimal cutoff levels on the dendrogram, which produced clusters capable of grouping together tweets associated with hate speech, cyber threats, or human threats, as well as benign topics.
590 $a School code: 1824.
650 4 $a Computer science. $3 199325
650 4 $a Web studies. $3 708690
650 4 $a Information science. $3 190425
653 $a Cyber threat
653 $a Natural language processing
653 $a Open source intelligence
653 $a OSINT
653 $a Twitter
653 $a Unsupervised clustering
690 $a 0984
690 $a 0646
690 $a 0723
710 2 $a Delaware State University. $b Computer Sciences. $3 886566
773 0 $t Masters Abstracts International $g 82-02.
790 $a 1824
791 $a M.S.
792 $a 2020
793 $a English
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=27957871