Welcome!

Virtualization Authors: Elizabeth White, Raja Patel, Liz McMillan, Dana Gardner, Adine Deford

Related Topics: Security, Java, SOA & WOA, Linux, Web 2.0, SDN Journal

Security: Article

Security Threats Continue to Grow

How Big Data and Machine Learning Can Work Together to Solve Security Threats

They read like a list of horror stories for businesses big and small alike. Sony’s PlayStation Network is hacked twice, exposing the personal information of 77 million customers. Zappos becomes the victim of a hack that exposes the addresses and phone numbers of 24 million people. Up to 81 million Yahoo email customers’ passwords are compromised, forcing the company to tell its users to reset them immediately. 110 million customers are affected when hackers infiltrate Target, and PIN numbers and credit card information are stolen. But these stories of major security breaches aren’t works of fiction--they actually happened, and it’s a concern businesses all over the world live with. Many companies are now turning to big data and machine learning as a way to tackle these risks and make sure valuable data is protected at all times.

Dealing with IT security issues is certainly nothing new for businesses. Computer viruses, malware, worms, and other threats have been around for a while, forcing companies to come up with solutions to either eliminate them or minimize the damages they cause. Much of this approach has been reactive in nature, essentially identifying a new threat or tactic hackers are using and developing the means to fight it. Older security systems had to search through smaller clusters of data to identify patterns that might indicate an attack, but the systems required significant resources and time to work, and even then their success rate was hit-and-miss. Systems were usually finding themselves being left behind by would-be attackers, forced to play catch-up in a game with a lot at stake.

With the growth of big data, data security has become even more complex and difficult to manage. More and more data is being created around the world, and trying to sort through all of it to identify security risks would tax older systems immensely. With new solutions desperately needed, many experts turned to machine learning. In simple terms, machine learning is a system that performs certain tasks by continuously learning from data without the need for specific programming. Machine learning can be used to detect security threats by sorting through all that data, something that simply wasn’t possible to that extent several years ago. Unlike traditional systems, which can get bogged down the more data they have to sort through, machine learning can actually get better if more data is added.

The way machine learning is able to detect security threats is by going through the data and identifying the signs and code that point to potential risks. This in turn creates a profile of what to look for, allowing machine learning and security systems to be able to predict and act on threats before they even happen. Essentially, machine learning can be used for security in much the same way it is used for advertising and marketing, targeting certain features it has determined through pattern recognition and using behavioral analytics to make more accurate predictions. This analysis is not only able to capture the hard data involved in security risks, it captures the context of risky events and can connect the relationships of those events to better understand just how threatening the risk actually is. This entire process takes less time than traditional systems and does not slow down productivity.

Threat detection through machine learning and big data was once out of reach for smaller businesses due to cost concerns and personnel requirements, but as these technologies have matured, smaller operations are now getting more access through big data cloud technology. The advances in recent years makes the utilization of machine learning possible for smaller security teams. In fact, security threat detection through machine learning is more of a hands-off process since machine learning systems undergo training on their own. The system is always learning, populating training sets to always get better at detecting security risks, even if they are new. The processing power and storage capabilities needed for machine learning are also within reach for small businesses thanks to advances in flash storage. The growing adaptability for companies makes security more robust and predictive instead of reactive.

There will never be a way to completely eliminate all security threats. Hackers and malware artists will always be looking for news ways to infiltrate and steal corporate information. But with a better understanding of the ways big data and machine learning can work together toward addressing this common problem, security breaches will be rarer and not as painful as those that have happened in recent years. A more secure future is definitely possible through machine learning.

More Stories By Gil Allouche

Gil Allouche is the Vice President of Marketing at Qubole. Most recently Sr. Director of Marketing for Karmasphere, a leading Big Data Analytics company offering SQL access to Apache Hadoop, where he managed all marketing functions, Gil brings a keen understanding of the Big Data target market and its technologies and buyers. Prior to Karmasphere, Gil was a product marketing manager and general manager for the TIBCO Silver Spotfire SaaS offering where he developed and executed go-to-market plans that increased growth by 600 percent in just 18 months. Gil also co-founded 1Yell, a social media ad network company. Gil began his marketing career as a product strategist at SAP while earning his MBA at Babson College and is a former software engineer.