SETTI: A <u>S</u> elf-supervised Adv <u>E</u> rsarial Malware De <u>T</u> ection Archi <u>T</u> ecture in an <u>I</u> oT Environment

In recent years, malware detection has become an active research topic in the area of Internet Things (IoT) security. The principle is to exploit knowledge from large quantities continuously generated malware. Existing algorithms practise available features for IoT devices and lack real-time prediction behaviours. More thus required on cope with misclassification input data. Motivated by this, this article, we propose adversarial self-supervised architecture detecting networks, SETTI, considering samples network traffic that may not be labeled. SETTI architecture, design three attack techniques, namely, Self-MDS , GSelf-MDS, ASelf-MDS . method considers data sample generation real-time. GSelf-MDS builds a generative model generate structure. Finally, utilises well-known perturbation techniques develop inject it over architecture. Also, apply defence mitigate these attacks, training, protect against injecting malicious samples. To validate algorithms, conduct experiments two datasets: IoT23 NBIoT. Comparison results shows dataset, most damaging consequences attacker’s point view reducing accuracy rate 98% 74%. NBIoT devastating algorithm can plunge 77%.