Machine learning in 5G IOT
Aim
Addressing the network management and security concerns in 5G powered Internet of things and Building a machine learning or deep learning algorithm for the same.
Introduction
There is an increasing number of 5G IoT systems, due to the advancement of electronics and communication techniques (e.g., wearable electronics, IoT devices, and 5G telecommunication solutions). Such technologies have enhanced the quality and performance of urban and suburban services, including healthcare, transport, energy, traffic, to name few.
In recent years, with the prevalence of 5G IoT systems, while AI technologies enable more autonomous and intelligent functions, the security of these systems has become more and more important as more and more personal data are generated and communicated through such modern 5G IoT systems.
Some of these emerging security problems cannot be solved by traditional security measures or by traditional privacy enhancement technologies. As a result,current 5G IoT system architectures are facing significant challenges to handle the security and privacy of increasing number of devices and servers as well as the protection of large volume of data that is processed in real-time.
Therefore, new security methods and privacy protection solutions which depend on deep learning are required to build more secure and better privacy-preserving 5G IoT systems. An increasing trend in integrating deep learning with access control, intrusion detection/prevention, and behaviour analysis of 5G IoT systemshas been recently observed.
Such integration will play a vital role in providing enhanced security for intelligent autonomous 5G IoT systems and enables organizations to make crucial changes to their security landscape.
Addressing the network management Problem in 5G IOT
Goal: Due to Rapid growth of number of devices the 5G IOT,The network is unable to cope up with this and eventually break down. The goal is to efficiently handle this using Machine learning and also prioritize the network based on needs.
Intution
One of the best things about 5G is the ability to implement virtual networks. These will create subnets that can have different traffic priorities. For example, in a hospital, the network could be designed to ensure that a connection between a surgeon and a robot was prioritized over, for example, communications being used by patients. Emergency transmissions can thus be protected even if the network is reaching capacity.
Proposed Solution
1.Tracking the consumption of every IP address in the network and allocating the network resources based on the intensity of Usage.
2.Desiging a parameter to find the highest intensity and it is given the priority when there is a congestion in the network.
3.Creating virtual subnets and ensuring that the ones with greater priorites are scattered among various subnets so that congestion would not occur in a particular subnet.
4.After doing all the above we use a Recurrent neural network to predict the network usage of a particular usage and hence allocation resources according to this data.
Concept:Time series prediction using LSTM neural networks
Here we consider the consumption of data every day at a particular time as a time series and then predict for the required time using LSTM
Prediction of Time series using LSTM is clearly explained in this article:https://machinelearningmastery.com/time-series-prediction-lstm-recurrent-neural-networks-python-keras/
Simulation
Let us consider a network in new york city. This data is taken from ISP provider in the new york city. Let us build a recurrent neural network which accepts the consumption of network as a time series data and predicts the usage of the network when the network gets congested.
Results of the predicted values and the simulation are depicted below
Link to the algorithm:https://github.com/avssridhar/5G-IOT-RESOURCE-MANAGEMENT/blob/master/5G IOT RESOURCE MANAGEMENT.ipynb
The code is clearly explained in the github repository above.
Addressing the privacy concerns in 5G IOT
Goal: 5G is undoubtedly one the very useful technologies when it comes to implementing IOT.But there are downsides as well.The security and privacy concerns are increasing day by day.The goal of the section is to build an effective machine learning algorithm which can be used to monitor the network and there by predict the attack.
Intution
When there is some attack on the network there would be some change in some of the network parameter.We also use the fack that hackers tresspass the network and there would be change in some parameter of the network and when the hacker comes again for the attack we would detect the anamoly in the network parameter and hence prevent the attack.We can use autoencode neural network for predicting the anamoly.
Proposed Solution
1.Use network error as the parameter for detecting anamoly
2.Find the network error everytime a new device enters the device.
3.Keep feeding this data to our autoencoder neural network algorithm.
4.Fix a threshold and if the anamoly value predicted due to the change in network error is greater than this threshold then block the IP address untill further authentication is done to that IP
This can be applied particularly to organizations to prevent attacks.All the userIP’s can be registered in the server and stored as data in the machine learning algorithm and the autoencoder can easily predict the anamoly when hacker attack the network.
Concept:Autoencoder Neural Network
Detailed explantion of autoencoder is given in this article:https://towardsdatascience.com/auto-encoder-what-is-it-and-what-is-it-used-for-part-1-3e5c6f017726
Simulation
Let us consider the data of an organization which underwent attacks and this data is obtained from kaggle.The algorithm rightly predicts the anamoly when there is an attack.
Refer this github repository for the code:https://github.com/avssridhar/intrusion-detection-in-an-iot-network/blob/main/INTRUSION DETECTION.ipynb
Hope you found the above explanation useful.Consider clapping for the article below. 😉
Feel free to reach out for further discussions.
Linkedin:https://www.linkedin.com/in/avs-sridhar-8b9904176/
Twitter:https://twitter.com/AvsSridhar2
