Microsoft Releases Open Source Machine Learning Security Framework

Machine learning has not only made great strides in fields such as finance, medical care, and military, but also has a great impact on the lives of citizens. However, Mircosoft said that 'although the attention to machine learning is increasing, cyber attack countermeasures against machine learning systems are insufficient', and the open framework ' Adversarial ML ' for companies that can detect, respond to and correct attacks on machine learning Threat Matrix has been released.

Cyberattacks against machine learning systems are more common than you think --Microsoft Security
https://www.microsoft.com/security/blog/2020/10/22/cyberattacks-against-machine-learning-systems-are-more-common-than-you-think/

Microsoft and MITER release framework to help fend off adversarial AI attacks | VentureBeat
https://venturebeat.com/2020/10/22/microsoft-and-mitre-release-framework-to-help-fend-off-adversarial-ai-attacks/

Microsoft has confirmed that cyber attacks targeting commercial machine learning systems have been increasing rapidly since around 2016, and Gartner, a major IT market research company, also said, 'Training data on cyber attacks using AI. attacks that target machine learning, such as pollution attacks that would 'reach 30% of the total by 2022 prediction has been announced. It can be said that the risk of cyber attacks related to machine learning is increasing day by day, but on the other hand, the sense of crisis about attacks on machine learning is not increasing (PDF file) According to a Microsoft survey , it was investigated 28 Twenty-five of the companies answered that they do not have security tools for machine learning systems.

To break through this situation, Microsoft is working with MITER, a non-profit research institute backed by the US government, which is also known for numbering the vulnerability identifier ' CVE ', to open source security for machine learning. The framework 'Adversarial ML Threat Matrix' has been released.

GitHub --mitre / advmlthreatmatrix: Adversarial Threat Matrix
https://github.com/mitre/advmlthreatmatrix

'Adversarial ML Threat Matrix' is a tool for enterprises intended for security personnel. It provides a framework for responding to new threats to machine learning systems, and adopts the same type of structure as the ' MITRE ATT & CK framework ' commonly used by security professionals. Microsoft and MITER have stated that they will examine the set of vulnerabilities and attacks in production and scrutinize the effectiveness of the Adversarial ML Threat Matrix.

Microsoft has also gathered opinions from experts at the University of Toronto, Cardiff University, Carnegie Mellon University, etc., saying that attacks on machine learning are an important research field. In addition, when the 'Adversarial ML Threat Matrix' was released, he said, 'We are looking for feedback and contributions regarding the framework,' and asked the community to provide information to be used to modify the framework.