DEVELOPMENT OF MODELS AND ALGORITHMS FOR DETECTING AND MITIGATING CYBERATTACKS IN INTERNET OF THINGS SYSTEMS

Abstract

This study focuses on the development of new algorithms and models
for detecting and mitigating cyberattacks targeting Internet of Things (IoT)
systems. The widespread use of IoT devices poses a serious threat to information
security, as most of these devices are resource-constrained and lack modern
security mechanisms. The paper carefully looks at the main dangers and kinds
of attacks in IoT systems (like DoS, spoofing, and sniffing), along with the
methods used to detect them, which include detection models and machine
learning algorithms. In particular, the advantages of artificial intelligence and
deep learning-based algorithms over traditional statistical approaches are
highlighted. We propose a hybrid approach for anomaly detection, network
traffic analysis, and real-time threat identification. The research utilizes a 100
GB dataset collected over 12 months from IoT devices. The proposed hybrid
model demonstrated a 27% improvement in accuracy compared to basic machine
learning methods, achieving an accuracy rate of 94.7%. Additionally, our model
reduced false-positive rates by up to 35% and increased real-time processing
speed by a factor of 2.3. The results of this research represent a major advance
in IoT security and introduce novel methods suitable for practical application in
industrial environments.