Using device-sourced data on quality and location, the millions of mobile devices running Curator Client software constantly gather information about a vast universe of Wi-Fi access points. But it is another Devicescape curation component, Curator Machine Learning Technology, that crunches this data over time and decides whether each hotspot should become or remain a part of the massive Devicescape Curated Virtual Network (CVN).
When new hotspots are identified, the machine learning technology attempts to validate them and record performance metrics and other characteristics. First, it checks whether the access points have already been ruled out in another recent evaluation. If not, they are then checked for adequate quality, access rules, keys, bypass methods, and authentication. If they pass these tests, they are added to the CVN and carefully monitored. Finally, if over the near term the hotspots show positive user experiences, they are retained in the CVN and monitored with each and every subsequent connection.
This is a very powerful and selective approach. Only the best-performing public access points—approximately 10 percent of those discovered—actually make the grade.
Devicescape's technology increases the accuracy of the curation process. Many parameters are checked and rechecked again and again. Some are weighed in context, such as level of sharing, which can logically be expected to show a considerably higher level at a large retail store than at a local corner store. If the data indicates very low intentional sharing, then it is likely the access point is an unprotected home network and thus not appropriate for the CVN.
Machine learning technology also increases the overall speed of the curation process. For example, the dynamic nature of the Devicescape approach dispenses with tedious techniques like war driving to gather SSIDs, or long waits for manual batch processing.
