Search

The Internet of Things (IoT) offers unprecedented opportunities to access anything from anywhere and at any time. It is, therefore, not surprising that the IoT acts as a paramount infrastructure for most modern and envisaged systems, including but not limited to smart homes, e-health, and intelligent transportation systems. However, the prevalence of IoT networks and the important role they play in various critical aspects of our lives make them a target for various types of advanced cyberattacks: Dyn attack, BrickerBot, Sonic, Smart Deadbolts, and Silex are just a few examples. Motivated by the need to protect IoT networks, this paper proposes SEHIDS: Self Evolving Host-based Intrusion Detection System. The underlying approach of SEHIDS is to equip each IoT node with a simple Artificial Neural Networks (ANN) architecture and a lightweight mechanism through which an IoT device can train this architecture online and evolves it whenever its performance prediction is degraded. By this means, SEHIDS enables each node to generate the ANN architecture required to detect the threats it faces, which makes SEHIDS suitable for the heterogeneity and turbulence of traffic amongst nodes. Moreover, the gradual evolution of the SEHIDS architecture facilitates retaining it to its near-minimal configurations, which saves the resources required to compute, store, and manipulate the model's parameters and speeds up the convergence of the model to the zero-classification regions. It is noteworthy that SEHIDS specifies the evolving criteria based on the outcomes of the built-in model's loss function, which is, in turn, facilitates using SEHIDS to develop the two common types of IDS: signature-based and anomaly-based. Where in the signature-based IDS version, a supervised architecture (i.e., multilayer perceptron architecture) is used to classify different types of attacks, while in the anomaly-based IDS version, an unsupervised architecture (i.e., replicator neuronal network) is used to distinguish benign from malicious traffic. Comprehensive assessments for SEHIDS from different perspectives were conducted with three recent datasets containing a variety of cyberattacks targeting IoT networks: BoT-IoT, TON-IOT, and IoTID20. These results of assessments demonstrate that SEHIDS is able to make accurate predictions of 1 True Positive and is suitable for IoT networks with the order of small fractions of the resources of typical IoT devices. [ABSTRACT FROM AUTHOR]

In this paper, a constructive one-hidden-layer network is introduced where each hidden unit employs a polynomial function for its activation function that is different from other units. Specifically, both a structure level as well as a function level adaptation methodologies are utilized in constructing the network. The functional level adaptation scheme ensures that the "growing" or constructive network has different activation functions for each neuron such that the network may he able to capture the underlying input-output map more effectively. The activation functions considered consist of orthonormal Hermite polynomials. It is shown through extensive simulations that the proposed network yields improved performance when compared to networks having identical sigmoidal activation functions. [ABSTRACT FROM AUTHOR]

In this paper, a proposed algorithm that dynamically changes the neural network structure is presented. The structure is changed based on some features in the cascade correlation algorithm. Cascade correlation is an important algorithm that is used to solve the actual problem by artificial neural networks as a new architecture and supervised learning algorithm. This process optimizes the architectures of the network which intends to accelerate the learning process and produce better performance in generalization. Many researchers have to date proposed several growing algorithms to optimize the feedforward neural network architectures. The proposed algorithm has been tested on various medical data sets. The results prove that the proposed algorithm is a better method to evaluate the accuracy and flexibility resulting from it. [ABSTRACT FROM AUTHOR]

The Internet of Things (IoT) offers unprecedented opportunities to access anything from anywhere and at any time. It is, therefore, not surprising that the IoT acts as a paramount infrastructure for most modern and envisaged systems, including but not limited to smart homes, e-health, and intelligent transportation systems. However, the prevalence of IoT networks and the important role they play in various critical aspects of our lives make them a target for various types of advanced cyberattacks: Dyn attack, BrickerBot, Sonic, Smart Deadbolts, and Silex are just a few examples. Motivated by the need to protect IoT networks, this paper proposes SEHIDS: Self Evolving Host-based Intrusion Detection System. The underlying approach of SEHIDS is to equip each IoT node with a simple Artificial Neural Networks (ANN) architecture and a lightweight mechanism through which an IoT device can train this architecture online and evolves it whenever its performance prediction is degraded. By this means, SEHIDS enables each node to generate the ANN architecture required to detect the threats it faces, which makes SEHIDS suitable for the heterogeneity and turbulence of traffic amongst nodes. Moreover, the gradual evolution of the SEHIDS architecture facilitates retaining it to its near-minimal configurations, which saves the resources required to compute, store, and manipulate the model’s parameters and speeds up the convergence of the model to the zero-classification regions. It is noteworthy that SEHIDS specifies the evolving criteria based on the outcomes of the built-in model’s loss function, which is, in turn, facilitates using SEHIDS to develop the two common types of IDS: signature-based and anomaly-based. Where in the signature-based IDS version, a supervised architecture (i.e., multilayer perceptron architecture) is used to classify different types of attacks, while in the anomaly-based IDS version, an unsupervised architecture (i.e., replicator neuronal network) is used to distinguish benign from malicious traffic. Comprehensive assessments for SEHIDS from different perspectives were conducted with three recent datasets containing a variety of cyberattacks targeting IoT networks: BoT-IoT, TON-IOT, and IoTID20. These results of assessments demonstrate that SEHIDS is able to make accurate predictions of 1 True Positive and is suitable for IoT networks with the order of small fractions of the resources of typical IoT devices.

In this paper, a proposed algorithm that dynamically changes the neural network structure is presented. The structure is changed based on some features in the cascade correlation algorithm. Cascade correlation is an important algorithm that is used to solve the actual problem by artificial neural networks as a new architecture and supervised learning algorithm. This process optimizes the architectures of the network which intends to accelerate the learning process and produce better performance in generalization. Many researchers have to date proposed several growing algorithms to optimize the feedforward neural network architectures. The proposed algorithm has been tested on various medical data sets. The results prove that the proposed algorithm is a better method to evaluate the accuracy and flexibility resulting from it.

It is well known that single hidden layer feedforward networks with radial basis function (RBF) kernels are universal approximators when all the parameters of the networks are obtained through all kinds of algorithms. However, as observed in most neural network implementations, tuning all the parameters of the network may cause learning complicated, poor generalization, overtraining and unstable. Unlike conventional neural network theories, this brief gives a constructive proof for the fact that a decay RBF neural network with n + 1 hidden neurons can interpolate n + 1 multivariate samples with zero error. Then we prove that the given decay RBFs can uniformly approximate any continuous multivariate functions with arbitrary precision without training. The faster convergence and better generalization performance than conventional RBF algorithm, BP algorithm, extreme learning machine and support vector machines are shown by means of two numerical experiments. [ABSTRACT FROM AUTHOR]