US Energy Department Floats Solution to Illicit Crypto Mining Malware
The DOE claims its detection software uses a deep-learning mechanism to identify cryptojackers, but it needs private-sector assistance selling the tool.
Scientists at the U.S. Department of Energy (DOE) are asking the private sector for help in commercializing a super-powerful cryptojacking detection algorithm that government officials believe can help datacenters overcome crypto-mining malware.
Disclosed but sparsely described in a Feb. 23 contract opportunity, the technology can sniff out illicit mining software, which harness hosts' spare computing power to mine cryptocurrencies with extreme accuracy. This detection software combats the "increasing threat" of burrowed cryptojacking malware, an expensive specter DOE said menaces data centers globally.
Indeed, cryptojackers have been caught hijacking data farms, government computers, major banks, medical research supercomputers, and hundreds of websites to mine crypto, netting their developers millions of dollars – often in privacy coins like monero. They cost their developers nothing but can drain electricity and computing resources from unwitting victim hosts.
DOE officials at the Idaho National Laboratory are now keen to head off cryptojackers by offering their detection technology to private sector partners.
Tech specs
Here's how the tech works in (relatively) simple terms: DOE's invention scans legitimate-seeming data processing applications for hidden cryptojackers that otherwise turn entire datacenters into zombie cryptocurrency mining farms. It spots these exploits reliably using a deep learning mechanism that researchers say is far more accurate than up/down vetting known as binary classification.
Here's how it works: "This invention is a rapid test based on machine translation to verify a binary submitted for execution on a datacenter system is free of cryptocurrency mining malware. It uses the attention mechanism in deep learning to accurately and reliably detect cryptocurrency malware surreptitiously deployed on high performance computing (HPC) systems. This approach is via machine translation rather than binary classification," the contract opportunity read.
In August, scientists at DOE's Los Alamos National Laboratory unveiled a neural network they said could detect cryptojackers faster and more reliably than non-AI methods.
"This type of software watchdog will soon be crucial to prevent cryptocurrency miners from hacking into high-performance computing facilities and stealing precious computing resources,” government researcher Gopinath Chennupati said at the time.
STORY CONTINUES BELOW
It was not immediately clear if the technology DOE now hopes to license out is related to the Los Alamos invention. DOE representatives did not immediately respond to CoinDesk queries.
The juicy details underlying DOE's invention rest behind a firewall that only prospective partners can access. But it is those corporate peers that DOE hopes to woo in an effort to develop and ultimately commercialize the early-stage cryptojacking detection algorithm.