Don’t be a Dumb Pipe! You CAN Analyze Encrypted Video Traffic

If you were to ask consumers in regulated countries what the main driver was for choosing their mobile service provider, chances are they’ll tell you that when it comes to data bundles and price, they’re all similar. So, what is most important to them?

Well here are a few hints.

By 2021, 82% of Internet traffic will be video.

78% of that video traffic is expected to be over mobile networks.

Bottom line… it all comes down to Quality of Experience (QoE). People will choose the service that lets them stream video to their mobile devices while they’re on the go, without interruptions.

For SPs looking to differentiate their service, it is critical to be able to analyze and maintain control over video traffic to ensure that they are maintaining the QoE that their subscribers expect.

Simple to do, right?

Well it is if your video is not encrypted.

But with video playing such a dominant part of data traffic, the major content providers are now encrypting video streams. Whether this is for reasons of security, commercial considerations, or search engine optimization, it doesn’t really matter. The fact is, where streaming video is concerned, service providers are blind to what goes on in their video streams and at risk of being turned into dumb pipes. As a result, it’s nearly impossible for them to analyze or influence QoE when their customers view encrypted video.

So, what can be done?

Objective measurements such as stalls, resolutions, and resolution changes could help you understand how your network is behaving in different regions. But these metrics are very difficult to capture from encrypted streams, and in any event, they don’t provide a complete picture of a customer’s quality of experience.

At Allot, we came up with a solution.

First, we developed AI-driven inline algorithms to detect encrypted video streams. Using machine learning techniques, we then derive the resolution, resolution changes, and stall metrics. By combining these metrics with additional parameters, and then applying further proprietary algorithms, we obtain a perceived QoE score.

That QoE score is then factored in with geographical, network topology, and time of day considerations to create actionable intelligence for service providers.

By mapping the perceived QoE scores to cell IDs through which the traffic passed on the mobile network, service providers can proactively detect where problems arise. These could be due to faulty equipment that requires replacing or high congestion areas that must be augmented with more bandwidth or network resources.

To that end, we created an Encrypted Video QoE Dashboard tool that helps SPs understand encrypted video QoE across their networks and gauge the most cost-effective ways to meet, and even exceed, their customers’ expectations.

Because the dashboard displays scores hourly, SPs can also identify peak periods of demand and so allocate more bandwidth during those hours.

Such visibility into encrypted video is rare, to say the least. As far as we know, Allot is the first DPI-based vendor to nail down YouTube QoE Analytics and turn it into actionable intelligence for SPs.

Given that YouTube makes up over 80% of video traffic streamed to Android devices in the US, that’s quite a game changer that will benefit SPs looking for the most cost-effective way to shape traffic without impacting QoE.

To see further examples of how CSPs can leverage actionable intelligence to improve end user QoE, read this Frost & Sullivan report.