Dlts: A Deep Learning Approach For Twitter Spam Detection

  • Sneha Prakash, Dr. R. Gunasundari
Keywords: DL, Twitter, Spam, DLTS, OSN, Extra Tree Classifier: ETC


In today's society, online social networking is rapidly expanding. Nonetheless, attacks against it are becoming more frequent; one such attack is the Twitter attack. This spammer disseminated various fraudulent tweets that may take the shape of links or hashtags on the website and online services, damaging legitimate users. However, existing methods do not properly and completely disclose whether or not Twitter is spamming. Spam tweets are detected and blocked by Google Safe Browsing and Twitter's Bot Maker tools, which help to curb spammers. These programs are capable of blocking harmful links. As a result, businesses and academics have taken several methods to create spam-free social network (SN) systems. Some are solely based on user-based features, while others are solely based on tweet-based features. We presented the DLTS technique for Twitter spam detection (TSD) to address this problem. The suggested approach integrates deep learning methods such as LR (Logistic Regression), Naive Bayes, Random Forest, CNN, DT, and ETC for effective spam identification. This method detects spammers by analysing both tweet content and user meta-data (e.g., account age, number of followers/followers, and so on). To assess the proposed method's performance on two distinct real-world datasets using two ML-based and two DL-based techniques, showing a performance improvement when using our methodology. The study's performance assesses detection accuracy, true positive/false positive, and the F measure; stability analyses algorithm using training samples of varying sizes. The goal of scalability is to understand better the effect of decreasing training time in learning algorithms.

How to Cite
Dr. R. Gunasundari , S. P. (2021). Dlts: A Deep Learning Approach For Twitter Spam Detection . Design Engineering, 2608-2620. Retrieved from http://thedesignengineering.com/index.php/DE/article/view/5176