If a user can be positively identified across multiple devices, for instance, because the user has logged into a platform such as Google, Facebook, Yahoo or Twitter, it can be “determined” who the user is for purposes of targeted advertising.
Probabilistic tracking depends upon collecting non-personal data regarding device attributes like operating system, device make and model, IP addresses, ad requests and location data, and making statistical inferences to link multiple devices to a single user. Note that this is accomplished through proprietary algorithms owned by the companies performing probabilistic finger-printing. Note also that in the EU IP Addresses are personal information.
Device graphs can be created by combining non-personal data regarding use of smartphones and other devices, (e.g., apps downloaded, operating systems, etc.) with personal log-in information to track interactions with content across multiple devices.