Cardiovascular disease is one of the most common diseases in the modern world, which, if diagnosed early, can greatly reduce the damage to the patient. Diagnosis of heart disease requires great care, and in some cases, the process can be disrupted by human error. Machine learning methods, especially data mining, have gained international acceptance in almost all aspects of life, especially the prediction of heart disease. On the other hand, datasets related to heart patients have many biological features that most of these features do not have a direct impact on diagnosis. By removing redundant features from the dataset, in addition to reducing computational complexity, the accuracy of heart patients’ predictions can also be increased. This paper presents a density-based unsupervised approach to the diagnosis of abnormalities in heart patients. In this method, the basic features in the dataset are first selected based on the filter-based feature selection approach. Then, the DBSCAN clustering method with adaptive parameters has used to increase the clustering accuracy of healthy instances and to determine abnormal instances as cardiac patients. Partition clustering methods suffer from the selection of the number of clusters and the initial central points and are very sensitive to noise. The DBSCAN method solves these problems by creating density-based clusters, but the selection of the neighborhood radius threshold and the number of connected points in the neighborhood remains unresolved. In the proposed method, these two parameters are selected adaptively to achieve the highest accuracy for the diagnosis and prediction of heart patients. The results of the experiments show that the accuracy of the proposed method for predicting heart patients is approximately 95%, which has improved in comparison with previous methods.

/pdf/anomaly-detection-in-heart-disease-using-a-density-based-15y8fceg.pdf

Anomaly Detection in Heart Disease Using a Density-Based Unsupervised Approach

Twitter bots amplify target content in a coordinated manner to make them appear popular, which is an astroturfing attack. Such attacks promote certain keywords to push them to Twitter trends to make them visible to a broader audience. Past work on such fake trends revealed a new astroturfing attack named ephemeral astroturfing that employs a very unique bot behavior in which bots post and delete generated tweets in a coordinated manner. As such, it is easy to mass-annotate such bots reliably, making them a convenient source of ground truth for bot research. In this paper, we detect and disclose over 212,000 such bots targeting Turkish trends, which we name astrobots. We also analyze their activity and suspension patterns. We found that Twitter purged those bots en-masse 6 times since June 2018. However, the adversaries reacted quickly and deployed new bots that were created years ago. We also found that many such bots do not post tweets apart from promoting fake trends, which makes it challenging for bot detection methods to detect them. Our work provides insights into platforms’ content moderation practices and bot detection research. The dataset is publicly available at https://github.com/tugrulz/EphemeralAstroturfing.

/pdf/analyzing-activity-and-suspension-patterns-of-twitter-bots-3op9kdi9.pdf

Analyzing Activity and Suspension Patterns of Twitter Bots Attacking Turkish Twitter Trends by a Longitudinal Dataset

To curb the problem of false information, social media platforms like Twitter started adding warning labels to content discussing debunked narratives, with the goal of providing more context to their audiences. Unfortunately, these labels are not applied uniformly and leave large amounts of false content unmoderated. This paper presents L AMBRETTA , a system that automatically identiﬁes tweets that are candidates for soft moderation using Learning To Rank (LTR). We run L AMBRETTA on Twitter data to moderate false claims related to the 2020 US Election and ﬁnd that it ﬂags over 20 times more tweets than Twitter, with only 3.93% false positives and 18.81% false negatives, outperforming alternative state-of-the-art methods based on keyword extraction and semantic search. Overall, L AMBRETTA assists human moderators in identifying and ﬂagging false information on social media.

LAMBRETTA: Learning to Rank for Twitter Soft Moderation

—Disinformation can be used to sway public opinion toward a certain political or economic direction, adversely impact public health, and mobilize groups to engage in violent disobedi- ence. A major challenge in mitigation is scarcity: disinformation is widespread but its mitigators are few. In this work, we interview fact-checkers, journalists, trust and safety specialists, researchers, and analysts who work in different organizations tackling problematic information across the world. From this interview study, we develop an understanding of the reality of combating disinformation across domains, and we use our findings to derive a cybersecurity-inspired framework to characterize the threat of disinformation. While related work has developed similar frameworks for conducting analyses and assessment, our work is distinct in providing the means to thoroughly consider the attacker side, their tactics and approaches. We demonstrate the applicability of our framework on several examples of recent disinformation campaigns.

/pdf/tactics-threats-targets-modeling-disinformation-and-its-3g6ypmcj.pdf

Tactics, Threats & Targets: Modeling Disinformation and its Mitigation

One out of four deaths is caused by heart related issues. Acting upon early signs of heart disease can, thus, drastically increase probability of saving lives. This paper discusses a cost-effective and reliable method of diagnosing heart abnormalities by using mobile phones that are nowadays typically available to an average user. A mobile application is developed to detect heart abnormal activities using either a digital stethoscope measurement as input, or a mobile recording of the heart beat using the mobile's microphone. To process the raw heart sound data, we first denoise the signal using wavelet transforms, and then apply machine learning techniques, namely, Convolutional Neural Networks for the classification of the stored heart sounds. A database consisting of recorded human heart sounds and their corresponding diagnosis is used to train the neural network. Moreover, neural network fine-tuning techniques such as ADAM Regularization is used to smoothen the prediction process. The proposed approach is tested on heart sound signals, that are 5 to 8 seconds long, and is shown to perform with an accuracy of 94.2% on the validation set.

Labeeba Begum

Papers

Tactics, Threats & Targets: Modeling Disinformation and its Mitigation

Detecting Heart Anomalies Using Mobile Phones and Machine Learning