Loading...
| Friend's email | |
| Your name | |
| Your email | |
| enter code | |
This page was sent successfuly
- Type of Document: M.Sc. Thesis
- Language: Farsi
- Document No: 56635 (05)
- University: Sharif University of Technology
- Department: Electrical Engineering
- Advisor(s): Hajsadeghi, Khosro
- Abstract:
- Nowadays, by integrating the Internet of Things systems into the daily life of humans, mankind has created a platform for providing numerous and diverse services through which life has become much simpler and more convenient. These systems have gradually become an integral part of today's life. They are used in many areas of production and service provision, such as healthcare, agricultural industry, supply chain, education system, transportation, and many others. Although these achievements have facilitated human life in many aspects, they are also associated with many security risks. Intrusion detection systems (IDS) are methods for predicting possible damage (through security attacks such as hacking and intrusion) in a network. Designing a desirable intrusion detection system in the Internet of Things networks using encryption is impossible due to the lack of data access permission and the ability to authenticate users. Therefore we are looking at the design of these systems differently. Designing an intrusion detection system based on machine learning (ML) technology can provide us with this capability. But in the meantime, the design of such systems will also bring challenges, including the low possibility of detecting the type of connection using a low amount of computing data and the lack of enough time to detect the type of attack. In this research, we will design a new model based on machine learning techniques and introduce a hybrid feature selection algorithm in order to overcome the introduced challenges. In this process, we implemented our proposed algorithm on the IOTID20 dataset precisely, and achieved an accuracy of 99.977% using the XGB algorithm. This result is obtained based on five input variables. Meanwhile, by introducing a new algorithm designed based on linear regression, we reached an accuracy of 98.541%, which was achieved when the model's training and testing time were 278 and 5 milliseconds, respectively. This result was the lowest training and testing time among the results of all other models and showed the superiority of our model over them
- Keywords:
- Machine Learning ; Internet of Things ; Intrusion Detection System ; Internet of Things Security
Digital Object List-
محتواي کتاب - view
Bookmark