Five Easy Pieces: How Machine Learning Is Already Boosting Cybersecurity

There are many good reasons why traditional security practices are becoming less effective at protecting against cyberattacks. There is too much security-related data flooding the network from an increasing number of users and devices. There is a lack of skilled personnel to watch over and analyze this data. And the security staff you have likely wastes too much time chasing down false positives. Valuable minutes — or even hours — can tick by before analysts and incident responders are aware of a threat.

Five Machine Learning Applications for Cybersecurity

Without being programmed or explicitly told what to do, machine learning analytics discover patterns leading to insights while creating predictive models that drive actions. The security industry uses various types of analytics to spot threats, including these increasingly popular machine learning techniques.

The truth is, machine learning and security are old friends, and we’ve all benefited from their harmony. For a long time, machine learning has cleansed virtually every inbox of spam by parsing the structure of emails, the words within them and images that might be placed therein. Then off into the spam dump go suspicious emails that may well carry malicious threats or phishing expeditions.

Here are five other, lesser-known ways machine learning helps security teams defend the enterprise:

1. Application Security

We’re applying machine learning to look at application source code. Subsequently, we try to identify patterns that are typical and in harmony with correct coding practices, learning about when vulnerabilities may be creeping into these applications. With machine learning, we can build stronger algorithms to root out those vulnerabilities.

2. User Behavior Analytics

User behavior analytics solutions look at users within your environment, then applies a set of learning algorithms that identify the typical patterns of usage per day, per application, per endpoint and so on. It identifies outliers indicating that someone has potentially compromised a user’s system by purloining his or her credentials. Machine learning detects this malfeasance while there is still time to minimize or even completely avoid the damage.

3. Data Security

Security analysts are regularly confounded by false positives, often with little or no means of assessing the real risks. Machine learning now is able to extend traditional database monitoring to probe far deeper into threat indicators based on relative changes in behavior.

Essentially, machine learning models the normal behavior patterns of users’ database activities based on historical data. It then analyzes new activities in-depth as they occur. Patterns of normal activities are instantly updated and any outliers are then detected in near real time before significant damage occurs.

4. Fraud Detection

Data scientists have made major progress deploying machine learning to detect fraud in the payments, insurance and other industries. Like other machine learning solutions, IBM Trusteer detects active threats while analyzing risk factors and flagging suspicious transactions. It is also available on an as-a-service basis to speed solution deployment. And unlike a lot of other cybersecurity spending, machine learning solutions aimed at fraud detection can actually lead to savings that drop to the bottom line.

5. Cognitive Analytics

Until recently, it wasn’t feasible to incorporate a large volume of important data, such as reports, blogs and emails, into security solutions. Watson for Cyber Security does precisely that. In doing so, it complements and augments the work of security analysts.

This solution leverages Watson’s machine learning-based natural language processing capabilities to ingest, analyze and report on tidal volumes of data. This is the same work an analyst might do manually, but Watson does so with unprecedented speed and scale. Cognitive security generates not only answers but also hypotheses, evidence-based reasoning and the ability to improve decision-making in real time.

Read the white paper: Outsmarting Fraudsters with Cognitive Fraud Detection

Share this Article:
Kevin Skapinetz

Director of Strategy and Product Design, IBM Security

Kevin Skapinetz is Director of Strategy and Product Design for IBM Security. He’s responsible for the overall strategic direction for IBM’s portfolio of products and services to protect customers from cyberthreats. Kevin recently led worldwide product marketing and over the last five years has been instrumental in building the IBM Security business unit into one of the largest security companies in the world. A 15-year veteran of the security industry, Kevin joined IBM from ISS (acquired by IBM in 2006), where he held multiple leadership positions in product management and engineering. He led product management for X-Force – one of the most respected commercial research organizations – and was chief strategist in the Office of the CTO, where he built the company's roadmap for securing emerging technologies, including virtualization and cloud computing.