Machine Learning and Enterprise Data Security

Companies are creating and handling more data than ever before. Keeping so much enterprise data secure has become increasingly challenging, even for organizations that have dedicated a lot of resources to it. Hackers have taken notice and are employing ever-more sophisticated attack strategies.

In order to successfully combat the ever-changing threat landscape, organizations must utilize the most cutting-edge security tool available: machine learning. Originally used almost exclusively for data analysis, machine learning is now an essential component of the most effective enterprise data protection strategies.

What Is Machine Learning?

Often used interchangeably with the term “artificial intelligence,” machine learning is, in fact, a separate branch of the larger field of AI. Machine learning consists in training and testing algorithms with datasets. In general, the more data available, the better trained an algorithm will be.

Ideally, the algorithm will learn from training data patterns that it will be able to apply to new data. For example, an algorithm trained on marketing data that includes the number of customers contacted and the number of goods purchased might identify a pattern between certain groups of consumers and those most likely to make a purchase.

A company can then use this information to create a more targeted marketing strategy based on the probability of a customer buying a good or service.

Machine learning is being applied in a number of industries because it is versatile and capable of providing a number of valuable and predictive insights, whether that means personalized marketing or enterprise data security.

The Data Landscape

In 2017, Symantec discovered 100 new malware families – a 36% global increase over the previous year. These data security loopholes didn’t go unnoticed, as more than 2 billion data points were compromised in the first half of 2017 alone, according to Gemalto.

While HIPAA legislation, the PCI Data Security Standard, and most recently GDPR, all provide strict rules and regulations concerning how companies must use, store, and protect consumer data, they do not apply to every industry or guarantee against a data breach.

The level of data security at the enterprise varies significantly from one company to the next. A Centrify and Dow Jones Customer Intelligence study reported that American corporations spent 86 billion on security in 2017. Yet, more than 2/3 of those surveyed had their data breached five times or more within the same period.

Today, 80% of all data is stored by companies. A report by CSC predicts that by 2020, enterprises will be managing a total of 28 zettabytes of data. One zettabyte is the equivalent of 1 billion gigabytes. That is an enormous amount of data to protect.

Data Breach Laws

Enterprise data includes both internally and externally communicated data. A breach within an organization, especially at the HR or administrative level could prove disastrous. Sensitive information about employees’ addresses, compensation, and government identification data could be discovered.

Likewise, an administrative level attack could give malicious agents a backdoor into the company’s customer data, financial transactions, and even development processes, any of which could cost a business millions.

Nearly every state requires companies to notify their customers in case of a data breach. Referred to as Personally Identifiable Information, this includes Social Security numbers, driver’s license numbers, and financial account numbers (e.g., bank/investment accounts) and, in some cases, medical data.

While states currently hold the power when it comes to prosecuting companies that violate these laws, many believe that federal legislation isn’t far behind.

Implementing a flexible, yet predictive data security strategy ahead of a federal mandate can help reduce the likelihood of a serious attack and or mitigate the fallout from one should it occur.

Machine Learning and Data Protection

Anti-malware software is usually signature-based; it identifies the unique digital fingerprint of a malicious program and then monitors specific devices to ensure that the same code doesn’t reappear. If it does, the software blocks it by preventing the code from executing.

But machine learning-based security systems work differently. Instead of looking for a specific pattern, they learn to identify the characteristics that make an event or action malicious. This is especially useful when trying to prevent attackers from compromising employee credentials, which is one of the most common and significant data security risks that enterprises face.

Machine learning is more flexible than traditional malware and more adept at successfully detecting a wide range of malicious threats. The threat landscape is constantly changing, and data protection strategies need to be able to keep up with attacks on operating systems, software and even processing chips (e.g., Spectre and Meltdown), all of which can wreak havoc on an enterprise’s ability to operate and maintain the trust of its clientele.

Elie Bursztein, who leads the anti-abuse team at Google says, “Before, we were in a world where the more data you had, the more problems you had. Now the more data the better…” She is referring to the capabilities of machine learning, specifically deep learning algorithms, which utilize neural networks that mimic the human mind, to not only train and evolve but also make independent judgments about each new data threat.

Google itself currently uses machine learning to detect payment fraud, safeguard its cloud service, and identify compromised systems.

The Defense Advanced Research Projects Agency (DARPA), a branch of the US Defense Department, is also using machine learning to protect data. By reverse-engineering malware, the agency is able to identify and categorize threats in real time.

A company with this technology can provide both its clients and its organization with unparalleled data security, while taking a proactive stance against the ever-changing threat landscape.

Security-based roadblocks can prevent even the most tech-savvy enterprise from growing and succeeding.

With machine learning-based data security, which, by necessity, must include multi-factor identification protocols, any company can quickly create a Zero Trust Security framework (ZTS) and scale it company-wide.

The ZTS model works by assuming that malicious agents already exist both inside and outside of every enterprise.

With trust no longer assumed, it must be demonstrated every time, whether someone is accessing a computer, a network or even a platform database. Strict though it may seem, this strategy significantly reduces the chances of a devastating and costly data breach.

Overview

Machine learning and data security are natural complements in a world where data breaches are a very common threat to enterprises and consumers alike. While machine learning can’t guarantee that a company will never be on the receiving end of a data breach, it can provide the kind of high-level data protection that can help keep an organization one step ahead of a potential attack.

Thanks Yaroslav Kuflinski, an AI/ML Observer at Iflexion. He has profound experience in IT and keeps up to date on the latest AI/ML research. Yaroslav focuses on AI and ML as tools to solve complex business problems and maximize operations.