A password for Netflix or HBO can be a precious commodity that a friend gives you, or that you offer to your partner (and we all save it on the computer to use it many times in the future). However, for the British company Synamedia, a shared password is a password that must be hunted. At CES (the world’s largest technology fair, held in Las Vegas) this year, the company introduced a new service that uses machine learning to detect shared passwords.
It works like this: a streaming service, like Netflix or HBO, buys access to Synamedia’s platform, which analyzes the data of all its users. It analyzes a variety of factors, such as where an account is accessed from, what time it is used, what content is viewed and on what device, etc. It then looks for patterns that indicate a shared password and gives the service provider a probability percentage, an assumption of how sure the system is of having found an attacker.
The system detects if the password is shared
A typical pattern would be to have a subscriber simultaneously viewing content in Madrid and A Coruña. Obviously, if this happens, it is quite unlikely that it is the same person who is seeing two things at the same time in different places. That is precisely what Synamedia is after , and what the most famous streaming platforms are willing to pay for, since in practice they are losing millions of euros in subscriptions.
After that, the service provider can choose what to do. If the sharing pattern is extreme, indicating that the passwords have been given away, or even sold over the Internet to multiple users, they may simply close your account. But if it’s something a little more innocuous, maybe a password shared between a family, they may just send you an email giving you a little slap on the wrist, and suggesting you upgrade to a premium account.
It’s all done by the algorithm
Machine learning is particularly well suited to these types of tasks because it can detect patterns in a large amount of data . More importantly, the consumption patterns that the algorithm studies are always evolving. What people watch and how they watch has changed tremendously in recent years. It’s better to have a system built on machine learning that can adapt to these changes, rather than a hard-coded algorithm that needs to be updated manually.
The data also reveals interesting patterns. Synamedia’s algorithms can detect university campuses, for example, as they tend to host peaks in consumers of online content through streaming platforms. That increased demand is evidence that the new streaming video market is maturing.
The new system is currently being tested by several companies, although Synamedia does not say which ones. However, the company sells other services to some of the biggest names in streaming.
