Statistical Detection of Online Drifting Twitter Spam

Liu, Shigang; Zhang, Jun; Xiang, Yang

doi:10.1145/2897845.2897928

Cited by 27 publications

(25 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In this field, there are also many methods that select account and/or tweet features as training data for the input of machine learning–based classifiers, such as the previous works . Account features can be the age of user account, the number of followings, and the number of followers.…”

Section: Related Workmentioning

confidence: 99%

“…Currently, many works have been proposed to detect spamming activities in Twitter or other social media platforms. Most current methods focused on establishing machine learning–based binary classifiers with statical features . These features could be selected from Twitter's streaming APIs and calculated by a JSON object.…”

Section: Introductionmentioning

confidence: 99%

“…However, because of the problems of spam drift and feature fabrication, the accuracy of the methods that adopted features statistics as the input to their classifiers was not satisfactory (F‐measure ≈ 87%). According to our previous study, Twitter spam drift problem was caused by the unpredictable features of newly posted tweets (ie, historical tweets cannot be used in prediction), and the feature fabrication problem was caused by the widely adopted social engineering techniques in Twitter spam activities. Spam features will be manipulated to be those which look very benign.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Detecting spamming activities in twitter based on deep‐learning technique

Wen

Liu

et al. 2017

Concurrency and Computation

Self Cite

View full text Add to dashboard Cite

Summary Twitter spam has long been a critical but difficult problem to be addressed. So far, researchers have developed a series of machine learning–based methods and blacklisting techniques to detect spamming activities on Twitter. According to our investigation, current methods and techniques have achieved the accuracy of around 87%. However, because of the problems of spam drift and information fabrication, these machine learning–based methods cannot efficiently detect spam activities in real‐life scenarios. Meanwhile, the blacklisting method also cannot catch up with the variations of spamming activities, as manually inspecting suspicious URLs is extremely timeconsuming. In this paper, we proposed a novel technique based on deep‐learning technique to address the above challenges. The syntax of each tweet will be learned through WordVector and trained by deep learning. We then constructed a binary classifier to differentiate spam and regular tweets. In experiments, we collected and labeled a 10‐day real tweet dataset as ground truth to evaluate our proposed method. We first went for empirical analysis with a series of comparisons to other methods: (1) performance of different classifiers, (2) other existing text‐based methods, and (3) nontext‐based detection techniques. According to the experiment results, our proposed method largely outperformed previous methods. We further conducted principle component analysis on typical methods to theoretically justify the outperformance of our method. We extracted all kinds of features via dimensionality reduction. It was found that our features were most distinct among all the detection methods. This well demonstrated the outperformance of our method.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Detecting spamming activities in twitter based on deep‐learning technique

Wen

Liu

et al. 2017

Concurrency and Computation

Self Cite

View full text Add to dashboard Cite

show abstract

“…1 Malicious URLs is one of the most commonly used methods that spammers implement malicious attacks. 2 Spammers usually share some interesting videos, stories, photographs, and information about discount, while these contents actually contain links to malicious websites. With the launching of short-link services in OSNs, spammers can take advantage of this short URL to hide the domain name of malicious website.…”

Section: Introductionmentioning

confidence: 99%

“…[2][3][4] The study demonstrates that a quite part of users will follow strangers who follow them in Sina Weibo. 5 There are some differences between the following strategies of normal users and spammers.…”

Section: Introductionmentioning

confidence: 99%

GroupFound: An effective approach to detect suspicious accounts in online social networks

Feng

Pan

et al. 2017

International Journal of Distributed Sensor Networks

View full text Add to dashboard Cite

Online social networks are an important part of people's life and also become the platform where spammers use suspicious accounts to spread malicious URLs. In order to detect suspicious accounts in online social networks, researchers make a lot of efforts. Most existing works mainly utilize machine learning based on features. However, once the spammers disguise the key features, the detection method will soon fail. Besides, such methods are unable to cope with the variable and unknown features. The works based on graph mainly use the location and social relationship of spammers, and they need to build a huge social graph, which leads to much computing cost. Thus, it is necessary to propose a lightweight algorithm which is hard to be evaded. In this article, we propose a lightweight algorithm GroupFound, which focuses on the structure of the local graph. As the bi-followers come from different social communities, we divide all accounts into different groups and compute the average number of accounts for these groups. We evaluate GroupFound on Sina Weibo dataset and find an appropriate threshold to identify suspicious accounts. Experimental results have demonstrated that our algorithm can accomplish a high detection rate of 86:27% at a low false positive rate of 8:54%.

show abstract

A comparative study of the class imbalance problem in Twitter spam detection

Liu

2017

Concurrency and Computation

Self Cite

View full text Add to dashboard Cite

Recently, online social network (OSN) such as Twitter has become an important and popular source for real-time information and news dissemination, and Twitter is inevitably a prime target of spammers. It has been showed that the security threats caused by Twitter spam can reach far beyond the social media platform itself. To mitigate the damage caused by Twitter spam, machine learning classification algorithms have been employed by researchers and communities to detect the Twitter spam. However, most of these studies have overlooked the class imbalance problem in Twitter spam detection. In this paper, we have studied the class imbalance problem in Twitter spam detection. Firstly, we have conducted a comparative study regarding some popular methods in handling the class imbalance problem in order to identify the most effective approach for addressing the class imbalance problem. Then, we have conducted another comparative study from Twitter spam detection based on several classic techniques. Experimental results demonstrate that a fuzy-based ensemble learning can significantly improve the classification performance on imbalance ground truth Twitter data. KEYWORDSclassification, class imbalance, online social network, Twitter spam detection INTRODUCTIONTwitter is used to exchange messages among friends. Unfortunately, spammers usually use Twitter as a tool to post unsolicited messages that contain malicious links, and even hijack trending topics. In this respect, the exponential growth of Twitter contributes to the increase of online spamming activities. Study shows that more than 3% messages are most probably abused by spammers. 1 In order to solve the security threats caused by spammers, a lot of researchers have proposed machine learning based algorithm for Twitter spam detection. However, most of these studies have neglected a fundamental issue that is the class imbalance problem, which is widely seen in real-world Twitter data. [2][3][4] The class imbalance problem has been identified as one of the ten challenging problems in data mining research. 5 This issue occurs in two different types of data sets: binary and multiclass. For binary problem, the training data from the minority class or positive class are very small, and the rest which make up the majority class or negative class are very large. While for multiclass problems, each of the class only contains a tiny fraction of samples. These problems are also especially critical in many real-world applications. For example, in Twitter spam detection, we used to have a large amount of normal Twitter data while only small number of spam samples, this gives us imbalanced data. Previous study has shown that the detection rate for Twitter spam can be decreased for about 33% in average with the class imbalance rate rises from 2 to 20. 6 Hence, a natural question in data mining research is how to improve the performance of classifiers facing with imbalanced data?Existing techniques for handling the class imbalance problem are mainly from three perspectives, includi...

show abstract

Statistical Detection of Online Drifting Twitter Spam

Cited by 27 publications

References 23 publications

Detecting spamming activities in twitter based on deep‐learning technique

Detecting spamming activities in twitter based on deep‐learning technique

GroupFound: An effective approach to detect suspicious accounts in online social networks

A comparative study of the class imbalance problem in Twitter spam detection

Contact Info

Product

Resources

About