MACHINE LEARNING FOR IOT DEVICE ANOMALY DETECTION ATTACK CLASSIFICATION

Abstract

 The theoretical underscores the security 
risks of Internet of Things (IoT) gadgets from 
programmers and assailants. IoT gadgets are 
defenseless against inconsistency assaults because of 
their interconnectedness. Support Vector Machine 
(SVM) and Random Forest (RF), stacking classifier, 
and voting classifier are utilized to recognize odd 
attacks in IoT gadgets in the Task. Every strategy is 
picked for its discovery and component determination 
abilities. The review explores different avenues 
regarding the arff NSL-KDD dataset. The suggested 
calculations RF and stacking classifier have great 
accuracy. Center around misleading positive rates 
shows a low rate in all cases. Featuring the technique's 
promising outcomes, quite Random Forests' preferable 
exactness over past writing. The stacking classifier and 
Random Forest exhibit promising exactness, review, 
and accuracy for recognizing and relieving IoT 
abnormal dangers. Troupe approaches like Voting 
Classifier (RF + AB) and Stacking Classifier (RF + 
MLP with LightGBM) consolidate various model 
expectations to make a more vigorous and precise 
forecast. Voting Classifier had 100 percent accuracy, 
Stacking Classifier 100 percent accuracy, and we 
made the front end involving flagon for client testing 
and IoT anomaly detection with client 
verification.[16]