Özet:

Abstract The world becomes an internet-based system as a result of the Internet of Things. IoT benefits people greatly in many different ways. The world is growing more technologically adept, yet a lot of problems are also concurrently emerging. Security is a primary priority considering that everything is connected. Occasionally, everyone hopes that their data will be transmitted safely over the internet. Currently, the rise of security concerns is proportionate to the development of technology. Due to the large number of vulnerable devices, IoT technologies are insecure and unreliable. IoT follows unique regulations and procedures, so conventional security measures cannot be applied. IoT security concerns need to be addressed on numerous fronts. Internet seclusion, wireless security, and privacy protection are only a few of them. Network intrusion and detection is a key topic of study in addition to the problems mentioned above. Due to the tens of thousands of networked devices in the IoT, it is extremely difficult to identify unusual access, unanticipated attacks, or odd device activity. In order to reduce security risks and detect intrusions in the Internet of Things, a number of tactics and algorithms have been proposed. Machine learning-based intrusion detection systems have demonstrated excellent accuracy and efficiency in recent years at spotting intrusions. This research offers a novel approach for boosting data transmission security, identifying attack-capable devices, and identifying devices that have been infiltrated by an intruder. The suggested approach uses Autoencoder (AE), a technique for machine learning that uses feature extraction along with principal component analysis. The website CloudStor is where you can find the intrusion and detection dataset. One unsupervised machine learning approach that effectively trains data is the autoencoder. The result shows that, when compared to other machine learning techniques, the suggested strategy produces better results.

Anahtar Kelimeler: