|
Briefing
|
- AI Breach Alert Tool – Cybersecurity company Darktrace and University of Cambridge researchers developing tool that can spot internal breaches using machine learning
- Unsupervised Learning – Uses unsupervised learning, type of machine learning where system is not fed with structured data to learn, unlike supervised learning, which needs human input
- Advantages – Allows machines to identify new anomalies in system, that do not necessarily follow trend of past attacks, with just few data points
- How It Works – Physical and digital sensors deployed around client’s network feed raw data to over 60 different algorithms that compete with one another to detect unusual activity, with master algorithm telling which algorithms to pay attention to and which to ignore
- Detection and Resolution – Final visualization allows human operators to see and resolve breaches by cutting all external communications from infected device
- Disclaimer – More sophisticated attacks can fool machines and go undetected as hackers try new methods
|
|
Accelerator
|
|
|
Sector
|
Information Technology
|
|
Organization
|
Darktrace Ltd., University of Cambridge
|
|
Source
|
|
|
Original Publication Date
|
November 16, 2018
|
