Medical Diagnosis Systems play a vital role in medical practice and are used by medical practitioners for diagnosis and treatment. In this paper, a medical diagnosis system is presented for predicting the risk of cardiovascular disease. This system is built by combining the relative advantages of genetic algorithm and neural network. Multilayered feed forward neural networks are particularly suited to complex classification problems. The weights of the neural network are determined using genetic algorithm because it finds acceptably good set of weights in less number of iterations. The dataset provided by University of California, Irvine (UCI) machine learning repository is used for training and testing. It consists of 303 instances of heart disease data each having 14 attributes including the class label. First, the dataset is preprocessed in order to make them suitable for training. Genetic based neural network is used for training the system. The final weights of the neural network are stored in the weight base and are used for predicting the risk of cardiovascular disease. The classification accuracy obtained using this approach is 94.17%.

Due to technology advancement, connectivity and intelligence have revolutionized the consumer electronics (CE) world and IoT systems pose a new cybersecurity threat because smart devices are constantly connected to sensors, which are connected directly to enormous cloud servers. To address the cybersecurity vulnerabilities associated with IoT in smart cities, we provide a Software Defined Network (SDN) orchestrated anomaly detection system based on Split Learning (SL). The traffic generated by the CE is pre-processed and the relevant traffic features are selected. Then the model is split into at least two sub-models using SL to detach model training from the necessity for direct access to raw data. Client-side sub-models are trained on local training data on client devices, whereas server-side sub-models are trained on servers. Experiments show that the proposed SL-based CE network anomaly detection framework decreases computation needs on the client side by just computing a sub-model rather than the complete computationally costly model. The proposed solution has the potential to detect compromised CE network traffic in smart city environments. As evidenced by our results, the proposed framework performed significantly better than the existing state-of-the-art anomaly detection system in terms of accuracy.

Denial of Service (DoS) attacks are prevailing as a significant threat in computer networks necessitating a system to detect DoS attacks for protecting the computing resources. The existing solutions to detect DoS attacks face the lacuna of dimensionality which escalates computational cost and false alarm rate. These issues have been addressed by proposing Class Scatter Ratio (CSR) and Feature Distance Map (FDM) based statistical approach for detecting DoS attacks. CSR calculates the weights of each feature for identifying the best by distance based classifier. FDM extracts the correlation among the features. The attack is detected by comparing the computed FDM of new traffic with normal and attack profile vectors. Three experiments were conducted and the performance evaluation reveals that the computational complexity, false alarm rate, and execution time are low for the proposed approach. Further, it is evident from the ten fold cross validations that the accuracy obtained by the proposed approach for all the datasets are within the 95% confidence interval. Moreover, the proposed statistical approach yields significant results compared to the existing feature selection and extraction techniques and state-of-the-art attack detection systems.