Follow

Machine Learning Overview

The video below covers the same content as this article.

Distil Networks is continually developing features and honing methods of detecting malicious users and requests targeting your site. However, a single nefarious user can cloak their requests using a variety of disguises, masquerading as multiple users, browsers, locations, and more to circumvent your security restrictions.

Distil has constructed request-based protections against threats in place, such as known violators, questionable referrers, and rate limits. However, to block actual malicious users—rather than merely their attempts—user-based protections hold the potential for a more comprehensive approach.

Distil’s machine learning feature answers this need. It adds yet another layer of protection to track and target destructive users by way of machine learning and behavioral modeling. 

What is machine learning?

Being a pattern recognition system, machine learning looks at all requests processed through Distil and parses out bad users. It focuses on client behavior over time, rather than single requests.

What does it look for?

Machine learning collects and analyzes a variety of user-based data, including:

  • Other locations the bot has been
  • How many IP addresses they’ve cycled through
  • How fast the bot is moving through content
  • How long they’ve been requesting access
  • Bot origin

Using an evolving behavioral model, this new Distil system progressively tweaks its profile of malicious users and sharpens its understanding of incoming threats.

Should I enable this protection?

Distil recommends machine learning be enabled for all customers. As a behavioral modeling tool, it becomes smarter and stronger as more data is tracked. Enabling machine learning protection does not affect the latency of your site.

How do I set in my machine learning threshold?

To protect your assets from suspected malicious users, your machine learning threshold is that point at which automated mechanisms are triggered. In general, Distil recommends thresholds be set less aggressive and gradually moved more aggressive. This reduces the chance of presenting unnecessary blocks or Captcha pages to valued users before the machine learning system has had an opportunity to learn over time.

Configure your machine learning settings directly in the Distil Portal.

Was this article helpful?
0 out of 0 found this helpful
Have more questions? Submit a request

Comments