Lightweight Machine Learning Classifiers of IoT Traffic Flows

摘要: IoT traffic flows have different from traditional devices statistics and their classification become an important task because of the exponentially growing number smart devices. Conventional Deep Packet Inspection systems that rely on inspection open fields in TLS DNS packets, trend encrypting makes machine learning based only viable option for future networks. Moreover, computational complexity models becomes crucial large-scale operations. In this work, we investigated whether simple models, such as Logistic Regression, SVM with linear kernel, a Decision Tree, suitable real-world deployments performance multiclass traces, given thoughtful features engineering. We introduced new flow feature categorical type describes set TCP-flag within flow. addition, removal correlated space transformation via PCA method showed usefulness terms prediction reduction. order to account online mode, limited maximal packets 10. estimate upper-bound features, compared algorithms Random Forest, Gradient Boosting feed-forward neural network. performed 4-fold cross-validation by metrics Accuracy F1-measure. The test results demonstrated increases F1-measure logistic regression 99.1% base case 99.6%, thus closely approaching more computationally expensive models. Overall, evaluation feasibility lightweight model practical deployment performance.