Data breaches are at an all-time high, because traditional cybersecurity methods just can’t predict human behavior. Learn how stateful machine learning delivers a robust, people-centric approach to cybersecurity! 

Copyright by

SwissCognitiveOver the past decade companies have been boosting their cybersecurity budgets and investments, yet data breaches are still on the rise. The reason? Human error. From breached customer and client records to phishing attacks that lead to compromised systems or direct wire fraud, human-layer issues are the number-one cause of data breaches. In fact, 88% of data breaches reported to the UK’s information commissioner’s office were caused by human factors.

“Organizations are really only as secure as the gatekeepers to these digital systems and data,” says Ed Bishop, co-founder and chief technology officer at Tessian. “No matter the industry or sector you might be in, if you have people controlling systems and data within your organization, you have human-layer vulnerabilities.”

These human layer attacks impact companies both financially and from a reputtion standpoint. For example, publicly traded companies suffer an average of a 7.5% drop in their share price after a data breach. It’s an instant reflection of loss of confidence, loss of reputation, and has an impact on business continuity going forward.

Why are employees your biggest vulnerability?

These breaches are increasing as the business world goes digital, employees are increasingly distributed, and email remains the main artery of communication for the human layer. It’s where some of the most sensitive information in an organization is shared, and yet there are still very few security checks in place. Ease of access to emails has only increased, with employees shooting out messages from laptops, smartphones, tablets, and now even watches. With the volume and speed of information transactions increasing, workers are simply more prone to making errors.

“Just think about how easy it is to misdirect an email when you’re in a rush, or how easy it is to click on a link from a sender that seems legit at a quick glance,” Bishop says. “Ultimately people are human and everyone makes mistakes, but until people start embracing that concept, this problem will just keep growing.”

Companies are simply too focused on protecting the machine layer, when it’s people that make up a company’s most important security layer. The only solution is to build technology that can protect company data by identifying and preventing attacks aimed at employees.

“The conversation has to move beyond blaming employees for accidental data loss or being prey to phishing attempts, to how can technology empower users to feel safe in their environment,” he explains.

Thank you for reading this post, don't forget to subscribe to our AI NAVIGATOR!


Where traditional cybersecurity methods fall short

Protecting company data requires a layered approach, in four parts: removing access to data and systems, adding security policies, boosting training and awareness, and adding a technological solution aimed at detecting and preventing human error.

Traditional cybersecurity methods rely broadly on rule-based technologies. This is great for capturing threats that can essentially be codified into if-this-then-that logic. For example, if the email says “internal only” in the subject and it’s getting sent externally, an algorithm can detect the breach and warn the user.

However rule-based approaches aren’t intelligent, can flag too often, and create too much noise. They’ll ultimately end up affecting the productivity and effectiveness of the employees that they’re trying to protect. And most importantly, they’re just not able to capture the kind of intricacies of human layer security problems.

The new human-layer security bridges the gap

Machine layer solutions are still essential. But human layer security is the natural next evolution for companies that are trying to innovate in the security space and expand their security protection.

“The reason traditional machine learning models can detect malware is because of the simple fact that malware is always malicious,” Bishop says. “However, with human layer security problems, this is no longer true.”

Everything with humans is dynamic and in flux, he explains. Relationships are formed during the duration of a project, and then they fall away. You worked with a counterpart a lot a year ago, but now it would be highly unusual for that counterpart to email you asking for an invoice to be paid. Traditional machine learning methods are ineffective at solving these human layer security problems, just because they don’t understand how relationships and scenarios change over time. of this concept that they need to understand time. To be effective, a machine learning solution needs to be able to say, at this exact moment in time, for this person and their relationships, does this behavior look unusual? That’s what stateful machine learning can do. […]