A deep learning approach for host-based cryptojacking malware detection

With the continued growth and popularity of blockchain-based cryptocurrencies there is a parallel growth in illegal mining to earn cryptocurrency. Since mining for cryptocurrencies requires high computational resource; malicious actors have resorted to using malicious file downloads and other methods to illegally use a victim’s system to mine for cryptocurrency without them knowing. This process is known as host-based cryptojacking and is gradually becoming one of the most popular cyberthreats in recent years. There are some proposed traditional machine learning methods to detect host-based cryptojacking but only a few have proposed using deep-learning models for detection. This paper presents a novel approach, dubbed CryptoJackingModel. This approach is a deep-learning host-based cryptojacking detection model that will effectively detect evolving host-based cryptojacking techniques and reduce false positives and false negatives. The approach has an overall accuracy of 98% on a dataset of 129,380 samples and a low performance overhead making it highly scalable. This approach will be an improvement of current countermeasures for detecting, mitigating, and preventing cryptojacking.