Over the past decades, cyber attackers have increasingly become experts at cracking systems and evading security measures. As a result, finding and accurately identifying malware is a critical challenge for many individuals and businesses worldwide.
Cyber-security experts have recently been looking into the potential of ML techniques for classifying malware and finding what actions should be taken to eradicate it. While some techniques displayed promising results, studies demonstrate that many of them can fail to accurately detect malware that they never experienced earlier.
In a bid to identify more dependable methods to classify malware, researchers at Orange Innovation Inc. recently undertook a study examining the potential of the quantum version of ML algorithms. The paper, pre-published on arXiv, provides some initial insights into the pros and cons of two types of quantum machine learning models, covering steps that could be examined in future cyber-security research.
The possibility to use AI for malware analysis is being examined since 2019, stated the co-author of the paper. The co-author along with a fellow researcher is taking up to explore what quantum technology can bring to this problem. With the mathematical background of the two researchers in two complementary areas, the theoretical knowledge is likely to serve to be an advantage to understand this subject.
Quantum machine learning could enable users to derive information from fewer data, opine the researcher duo. Meanwhile, in order to test this hypothesis for malware classification, so far, they assessed the performance of two types of quantum machine learning models, known as QNN and QSVM.
“QSVM is the first algorithm that was tested. It is an adaptation of the Support Vector Machines algorithm in quantum,” explained one of the researchers. This was followed by testing QNN, a quantum adaptation of a classical neural network.
