Skip to main content
CoinPolice: Detecting Hidden Cryptojacking Attacks with Neural Networks
  1. publications
  2. anti-abuse

CoinPolice: Detecting Hidden Cryptojacking Attacks with Neural Networks

Available Media Publication (PDF)
Conference arXiv preprint (arXiv) - 2020
Authors Ivan Petrov , Luca Invernizzi , Elie Bursztein
Citation BibTeX
BibTeX
@inproceedings{Petrov2020CoinPolice,
  title = {CoinPolice: Detecting Hidden Cryptojacking Attacks with Neural Networks},
  author = {Ivan Petrov and Luca Invernizzi and Elie Bursztein},
  booktitle = {arXiv preprint},
  year = {2020},
  organization = {arXiv}
}

Cryptojacking uses a visitor’s computing resources to mine cryptocurrency without their consent. Hidden miners can evade common defenses through obfuscation, WebAssembly, changing domains and throttling their activity.

CoinPolice varies the CPU resources available to a browser and studies how the workload responds. A neural classifier uses that behavior to identify concealed mining, including aggressively throttled miners. The paper reports a 97.87% detection rate with a 0.74% false positive rate and applies the method in a large-scale investigation that identifies 6,700 mining sites.

newsletter signup
newsletter signup