Continuous authentication using deep neural networks ensemble on keystroke dynamics

During the last years, several studies have been proposed about user identification by means of keystroke analysis. Keystroke dynamics has a lower cost when compared to other biometric-based methods since such a system does not require any additional specific sensor, apart from a traditional keyboard, and it allows the continuous identification of the users in the background as well. The research proposed in this paper concerns (i) the creation of a large integrated dataset of users typing on a traditional keyboard obtained through the integration of three real-world datasets coming from existing studies and (ii) the definition of an ensemble learning approach, made up of basic deep neural network classifiers, with the objective of distinguishing the different users of the considered dataset by exploiting a proper group of features able to capture their typing style. After an optimization phase, in order to find the best possible base classifier, we evaluated the ensemble super-classifier comparing different voting techniques, namely majority and Bayesian, as well as training allocation strategies, i.e., random and K-means. The approach we propose has been assessed using the created very large integrated dataset and the obtained results are very promising, achieving an accuracy of up to 0.997 under certain evaluation conditions.

Machine Learning Approaches for Quality Assessment of Protein Structures

Protein structures play a very important role in biomedical research, especially in drug discovery and design, which require accurate protein structures in advance. However, experimental determinations of protein structure are prohibitively costly and time-consuming, and computational predictions of protein structures have not been perfected. Methods that assess the quality of protein models can help in selecting the most accurate candidates for further work. Driven by this demand, many structural bioinformatics laboratories have developed methods for estimating model accuracy (EMA). In recent years, EMA by machine learning (ML) have consistently ranked among the top-performing methods in the community-wide CASP challenge. Accordingly, we systematically review all the major ML-based EMA methods developed within the past ten years. The methods are grouped by their employed ML approach—support vector machine, artificial neural networks, ensemble learning, or Bayesian learning—and their significances are discussed from a methodology viewpoint. To orient the reader, we also briefly describe the background of EMA, including the CASP challenge and its evaluation metrics, and introduce the major ML/DL techniques. Overall, this review provides an introductory guide to modern research on protein quality assessment and directions for future research in this area.