Cybersecurity Approaches to Cryptojacking Detection: A Computer Application Perspective

Cryptojacking is the covert use of computing resources for cryptocurrency mining; it is a rising threat to a variety of different environments.This paper compares three supervised classifiers, i.e., Random Forest (RF), Support Vector Machine (SVM, RBF kernel), and Multilayer Perceptron (MLP), for the cryptojacking detection on a public dataset of 95,312 time-stamped observations (58 numeric features).An anti-leakage preprocessing pipeline is enforced, and models are evaluated under stratified random splitting, chronological splitting, cross-validation, and rolling-origin time validation.Under random splitting, all classifiers get near-perfect accuracy (RF: 1.0000, SVM: 0.9999, MLP: 1.0000).Under chronological splitting, both RF and MLP maintain their performance, while SVM shows a decrease in recall compared to before, highlighting the importance of time-aware validation.All errors are false negatives only.Area Under the Precision-Recall Curve (PR-AUC) and a majority class baseline are reported to deal with the class imbalance.RF feature importance analysis shows that memory and filesystem are the key metrics in classification, which can be used to guide lightweight monitoring.Efficiency benchmarks are supplied to aid in deployment-focused model selection.