This article presents a novel data encryption technique suitable for Internet of Things (IoT) applications. The cryptosystem is based on the application of a Catalan object (as a cryptographic key) that provides encryption based on combinatorial structures with noncrossing or nonnested matching. The experimental part of this article includes a comparative analysis of the proposed encryption method with the Catalan numbers and data encryption standard (DES) algorithm, which is performed with machine learning-based identification of the encryption method using ciphertext only. These tests showed that it is much more difficult to recognize ciphertext generated with the Catalan method than one made with the DES algorithm. System reliability depends on the quality of the key, therefore, statistical testing proposed by National Institute of Standards and Technology was also performed. Twelve standard tests, the approximate entropy measurement, and random digression complexity analysis are applied in order to evaluate the quality of the generated Catalan key. A proposal for applying this method in e-Health IoT is also given. Possibilities of applying this method in the IoT applications for smart cities data storage and processing are provided.
This paper considers a class of hybrid (heterogeneous) ensembles purely composed of symbolic elements. In learning diagnostic rules from gene expressions they demonstrate a significant improvement of accuracy with a small number of ensemble elements. This makes them suitable for learning of understandable knowledge, leading to diagnosis and its explanation in original terms (attributes).
In this paper we consider a direct application of advanced machine learning methods to standard model of time sequences data to avoid preprocessing. Besides classical machine learning methods, such as support vector machines, we used kernel learning to improve accuracy of learned knowledge. Kernel learning, especially multiple kernel learning (MKL), allows automated model creation to describe complex data and performs feature selection. The approach is tested using several publicly available machine learning software tools and time series datasets and its good generalization properties are demonstrated.