Hybrid network intrusion detection system for IoT attacks

A method for intrusion detection in an internet of things (IoT) environment, comprising:

• evaluating with a computer or system of computers sensed IoT data which performs feature selection based on information gain and which produces a plurality of features;
• extracting with the computer or system of computers dendritic cell algorithm signals from the plurality of features using a self normalizing neural network and categorizing each as being safe or dangerous,
• wherein a sigmoid activation function at output neurons assures that the signals are contained in a range 0-100,
• wherein a scaled exponential linear units (SeLU) activation function at hidden layer neurons in the self normalizing neural network cuts off high or low values of a feature attribute depending on signs of parameters vS,i or vD,i,
• wherein positivity restriction on weights wS,i and wD,i breaks a symmetry between formulae for safe signals and danger signals, and
• wherein the self normalizing neural network is prevented from choosing a solution where wS,i=−wD,i;
• and
• processing with the computer or computer system the extracted signals using a deterministic dendritic cell algorithm to identify intrusion detection.