Compromised account detection using authorship verification: a novel approach

Compromising legitimate accounts is a way of disseminating malicious content to a large user base in Online Social Networks (OSNs). Since the accounts cause lots of damages to the user and consequently to other users on OSNs, early detection is very important. This paper proposes a novel approach based on authorship verification to identify compromised twitter accounts. As the approach only uses the features extracted from the last user's post, it helps to early detection to control the damage. As a result, the malicious message without a user profile can be detected with satisfying accuracy. Experiments were constructed using a real-world dataset of compromised accounts on Twitter. The result showed that the model is suitable for detection due to achieving an accuracy of 89%.

A novel method for recommendation systems using invasive weed optimization

One of the popular approaches in recommendation systems is Collaborative Filtering (CF). The most significant step in CF is choosing the appropriate set of users. For this purpose, similarity measures are usually used for computing the similarity between a specific user and the other users. This paper proposes a new invasive weed optimization (IWO) based CF approach that uses users' context to identify important and effective users set. By using a newly defined similarity measure based on both rating values and a measure values called confidence, the proposed approach calculates the similarity between users and thus identifies and filters the most similar users to a specific user. It then uses IWO to calculate the importance degree of users and finally, by using the identified important users and their importance degrees it predicts unknown ratings. To evaluate the proposed method, several experiments have been performed on two known real world datasets and the results show that the proposed method improves the state of the art results up to 15% in terms of Root Mean Square Error (RMSE) and Mean Absolute Error (MAE).

Recommendation system using a deep learning and graph analysis approach

When a user connects to the Internet to fulfill his needs, he often encounters a huge amount of related information. Recommender systems are the techniques for massively filtering information and offering the items that users find them satisfying and interesting. The advances in machine learning methods, especially deep learning, have led to great achievements in recommender systems, although these systems still suffer from challenges such as cold-start and sparsity problems. To solve these problems, context information such as user communication network is usually used. In this paper, we have proposed a novel recommendation method based on Matrix Factorization and graph analysis methods, namely Louvain for community detection and HITS for finding the most important node within the trust network. In addition, we leverage deep Autoencoders to initialize users and items latent factors, and the Node2vec deep embedding method gathers users' latent factors from the user trust graph. The proposed method is implemented on Ciao and Epinions standard datasets. The experimental results and comparisons demonstrate that the proposed approach is superior to the existing state-of-the-art recommendation methods. Our approach outperforms other comparative methods and achieves great improvements, i.e., 15.56% RMSE improvement for Epinions and 18.41% RMSE improvement for Ciao.