Intrusion Detection System for Internet of Things-An Enhanced model
As the world is adapting to new technologies and the devices which are connected to the internet are growing enormously day by day. It is well known that the Internet of Things (IoT) is considered as a vital platform for sharing resources and deals with large data. Now, the major concern is how to protect the data in the IoT environment from the Intruders or Hackers as the devices in the IoT environment are connected through the internet and more prone to a vulnerability. There are various models which discussed on securing the data on IoT environment and Intrusion detection System is one among them.
Now we have to know which methodology is apt for making Intrusion detection system to detect intrusion with more accuracy and precision. The machine learning algorithm known as an Artificial Neural Network is used to develop a better network architecture. Now, we know that from the previous works performed on IDS, ANN is best to give accurate output and can detect the intruder with more accuracy.
Artificial Neural Networks Overview
The above figure shows that the Artificial Neural Network contains 3 layers
1) Input Units
2) Hidden Units
3) Output Units
The ANN classifier contains Hidden layers and inputs and the data need to be fed into it from the database. Now the main concept is to train the data and to train the hidden units the data is selected optimally by using SHO (Spotted Hyena Optimization Algorithm).
SHO Overview:
In the recent era, The SHO algorithm proved to be providing better results for many problems of the optimization. We will try to apply this algorithm to train the hidden units of ANN classifiers so as to enhance the minimum error and maximum accuracy.
The main steps involved in SHO are
1) Searching for the prey
2) Encircling the prey
3) Attacking the prey
In detail, we can say that it starts with generating the initial spotted hyena population and then calculating the fitness value of each search agent by using the fitness function formula. After that we have to define the set of optimal solutions. Later, need to apply SHO algorithm and update the position of each search agent and again repeat the step to find the fitness value and update the position till we get the optimal solution with minimum error or maximum accuracy.
The overall analysis should be carried out by the evaluation metrics which are necessary to calculate and analyze for the data from datasets are as follows
Positive Measures (This value we should get high when compared with other algorithms)
· Accuracy
· Precision
· Sensitivity
· Specificity
· Negative predictive value
· F1-score
· Matthews correlation coefficient
Negative Measures (This value we should get it low when compared to the results of other algorithms)
· False positive rate
· False negative rate
· False Discovery rate
In order to make it more enhanced model and detect the threat with more accuracy we need to have an improvised training data for that we can use Spotted Hyena based Optimization Algorithm for IoT Framework.
