Your browser is not supported. Please update it.

18 August 2017

AWS launches Amazon Macie, for ML-based data enterprise leak prevention

Amazon Web Services has launched Amazon Macie, a new machine-learning tool for identifying sensitive information inside an AWS deployment, in order to better protect it, and for monitoring anomalies in data access. With data-leaks being something of a constant headline in the technology sector, AWS is hoping that Macie helps its enterprise customers help themselves – and not leave backdoors open.

Many of the largest enterprises will go through a period of flux, as they on-board IoT devices and applications to improve their margins – which will both add a huge pile of data to their existing IT infrastructure, as well as greatly increasing their attack footprint. For many of these customers, there will be a sharp difference between the types of data they now have stored in their clouds, but AWS is hoping to intervene on their behalf – to prioritize the safe-keeping of the most sensitive data.

An example of this might be a hospital, which is looking to add smart building automation to improve its operational efficiency, while simultaneously connecting more of its healthcare equipment. There’s a great deal of difference between the on-off status of connected lighting or the temperature of a room and the sensitive information generated by something like a medical monitor that is feeding readings into a digital patient health record.

In that instance, an AWS customer might not care about the security of the smart building data, but a breach of the patient record could be disastrous – both for the privacy of the patient/s, as well as the resulting punishment from the associated government regulator. Because of the disparate nature of the two deployments, the hospital’s IT department or provider might have no idea that they have somehow exposed the data, and this is where AWS is hoping to intervene.

We can’t help but think that this is part of a strategy from AWS to help its brand stop appearing in coverage of new data leaks – when its customers do dumb things with sensitive data on the AWS platform, such as the recent Verizon, OneLogin, and Chicago breaches. As the largest cloud computing provider, its name is always going to be close to the story of the next big leak or outage. Even though the user at fault, AWS wants to help reduce the number of issues, so that its brand and reputation aren’t tarnished.

Macie originated in Harvest.ai, a company that AWS acquired in January that was founded by two former NSA staffers, which has had more features added to it since. TechCrunch reports that AWS paid $19m for Harvest.ai, a company comprised of 12 staff that had raised $2.3m at the time – a nice return for its investors.

So now AWS is putting Macie to use, and looks like the tool is more fleshed-out than Microsoft’s Azure SQL database Threat Detection, or Google Cloud’s Data Loss Prevention API. To be clear, AWS isn’t being particularly revolutionary here, and its rivals are able to offer a similar service, but it’s interesting to see the way that it is being pitched at enterprises – a small purchase that can greatly improve data security.

To be able to sift through the incredible amount of data stored on its servers, AWS has turned to the machine-learning functions of Macie – to speed up the process of hunting for insecure data in its Simple Storage Service (S3). AWS says that support for its other data storage offerings will arrive later this year, but Macie is initially confined to S3 deployments in its US data centers.

The system has been built to recognize things like personally identifiable information (PII – banking details, social security numbers, etc.) and intellectual property (IP), and displays alerts via a dashboard. This logs how that potentially vulnerable data is accessed or moved, and will use the machine-learning functions to spot patterns or anomalies that indicate risk. When Macie thinks it has spotted unauthorized access or an accidental leak, the alarm bells start ringing.

There’s no way that such a system could work on a manual intervention and inspection basis – there aren’t enough man-hours to be found for a price low enough to make this an enticing service for customers. While automated software could be used to a similar end, AWS is hoping that Macie’s machine-learning functions help it work in a more intelligent manner – spotting the signal amidst the noise, and actually learning from the data it handles, improving its natural language processing.

And the signal-noise ratio is important for the alerts, as a ‘dumb’ but automated software approach could throw up so many alerts (sysadmin working from home that day, CIO accesses a dashboard while on holiday, etc.) that the customer ends up either ignoring them or canceling the service out of frustration.

AWS lists Autodesk (design and CAD software house), Edmunds (car reviews and comparison website), and Netflix (OTT video titan) as flagship customers, with Autodesk saying it helped with security and compliance, and Edmunds noting that Macie’s granular level of inspection and intelligence provided it with “a completely new level of confidence in the security of our infrastructure.” Netflix said the service was used for tasks that previously required custom coding, in a flexible manner.

In terms of paying for Macie, customers are charged for the gigabytes of data that it classifies – after activating Macie from their AWS Management Console. Because of the classification process and workload, the payment for the first month will be much higher than following months, as Macie passes its eyes over the S3 Buckets of data. Once it has categorized everything, only the new data needs to be processed, so the costs fall sharply. AWS’ example pricing costs $533 for that first month, processing 106GB of data in a 1.5TB deployment, but then drops to $38.30 in the second month, and $38.60 in month-three.

“When a customer has a significant amount of content store in Amazon S3, identifying and classifying all of the potentially sensitive data can feel a bit like finding needles in a very large haystack – especially when monitoring tools that aren’t smart enough to effectively automate what is now a very manual process,” said Stephen Schmidt, AWS’ Chief Information Security Officer.

Schmidt added that “Amazon Macie approaches information security in a more intelligent way. By using machine-learning to understand the content and user behavior of each organization, Amazon Macie can cut through huge volumes of data with better visibility and more accurate alerts, allowing customers to focus on securing their sensitive information, instead of wasting time trying to find it.”