Design of Effective Spam Account Detection Techniques in Social Networks by Utilizing Survival Analysis and Auto Encoder Models

Abstract

Cloud computing is an emerging technology involved in the provision of on-demand and scalable service features. At present, cloud technology is highly utilized for big data storage for intensive applications. Thus, data security and privacy is a major concern for different applications based on the data level. However, security and privacy in the cloud need to be maintained and processed throughout the cycle from generation, use, transfer, archival, and storage. Social media platform comprises the huge data volume shared between users. Due to the presence of huge data volume security schemes need to be implemented with adequate availability, integrity, privacy, and confidentiality for data stored in the cloud. newlineThe present research concentrated on the construction of the data security model for social media. The social media sites are subjected to limitations in terms of nature, shortened form, size, and slang it is difficult to identify the spam account. This research concentrated on the spam account detection in the Twitter dataset for the classification of original and spam accounts. To perform spam account detection and classification this research is adopted in three stages. In the first stage, the Twitter data set is processed and estimated for the collected Twitter data for processing. To evaluate the collected Twitter data survival analysis is performed. The survival analysis is performed for evaluation of the available data for the processing. Secondly, the autoencoder is applied over the survival analysed data for the estimation to perform classification and detection. Finally, the RNN model is applied for Twitter spam account detection and classification of the data. The proposed model is stated as AeRNN for spam account detection. newlineThe presented work concentrated on the Twitter spam and non-spam account classification through an auto-encoder-based model. The developed model comprises the survival analysis model for the computation of the data attributes. The survival analysis model is applied over the auto-encoder model for processing a large volume of Twitter data to perform dimensionality reduction. The classification of the spam and non-spam account is estimated based on the RNN model. The RNN network comprises the different layers to process the Twitter data for the classification of spam and non-spam account. newlineThe developed AeRNN model uses the survival analysis for estimation of the feature vector of the social media data. Estimated feature attributes are computed for estimation and identification of the spammer account. In the auto-encoder model, the collected Twitter data were processed in vector form, and dimensions are minimized. The classification is performed with the RNN model for spam and non-spam account. The experimental analysis of the developed model is comparatively examined with the conventional technique. The analysis expressed that the developed model exhibits significant performance than the conventional techniques for the classification of spam and non-spam account on Twitter. newline