Abstract-Keystroke dynamics is one of the authentication mechanisms which uses natural typing pattern of a user for identification. In this work, we introduced Dependence Clustering based approach to user authentication using keystroke dynamics. In addition, we applied a k-NN-based approach that demonstrated strong results. Most of the existing approaches use only genuine users data for training and validation. We designed a cross validation procedure with artificially generated impostor samples that improves the learning process yet allows fair comparison to previous works. We evaluated the methods using the CMU keystroke dynamics benchmark dataset. Both proposed approaches outperformed the previous state-of-the-art results for the CMU dataset for unsupervised learning.

With the emergence of cloud computing, many attacks, including Distributed Denial-of-Service (DDoS) attacks, have changed their direction towards cloud environment. In particular, DDoS attacks have changed in scale, methods, and targets and become more complex by using advantages provided by cloud computing. Modern cloud computing environments can benefit from moving towards Software-Defined Networking (SDN) technology, which allows network engineers and administrators to respond quickly to the changing business requirements. In this paper, we propose an approach for detecting application-layer DDoS attacks in cloud environment with SDN. The algorithm is applied to statistics extracted from network flows and, therefore, is suitable for detecting attacks that utilize encrypted protocols. The proposed detection approach is comprised of the extraction of normal user behavior patterns and detection of anomalies that significantly deviate from these patterns. The algorithm is evaluated using DDoS detection system prototype. Simulation results show that intermediate application-layer DDoS attacks can be properly detected, while the number of false alarms remains very low.

We propose an ensemble clustering approach using group diffusion to reveal community structures in data. We represent data points as a directed graph and assume each data point belong to single cluster membership instead of multiple memberships. The method is based on the concept of ensemble group diffusion with a parameter to represent diffusion depth in clustering. The ability to modulate the diffusion-depth parameter by varying it within a certain interval allows for more accurate construction of clusters. Depending on the value of the diffusion-depth parameter, the presented approach can determine very well both local clusters and global structure of data. At the same time, the ability to combine single outcomes of the method results in better cluster segmentation. Due to this property, the proposed method performs well on data sets where other conventional clustering methods fail. We test the method with both simulated and real-world data sets. The results support our theoretical conjectures on improved accuracy compared to other selected methods.

Abstract-This article proposes a scalable version of the Dependence Clustering algorithm which belongs to the class of spectral clustering methods. The method is implemented in Apache Spark using GraphX API primitives. Moreover, a fast approximate diffusion procedure that enables algorithms of spectral clustering type in Spark environment is introduced. In addition, the proposed algorithm is benchmarked against Spectral clustering. Results of applying the method to real-life data allow concluding that the implementation scales well, yet demonstrating good performance for densely connected graphs.

No abstract