Systems and methods for distinguishing human users from bots

One or more computer-readable non-transitory storage media embodying software that is operable when executed to:

• identify, based on predetermined criteria, a first set of users of an online social-networking system as belonging to a first population segment and a second set of users of the online social-networking system as belonging to a second population segment;
• in response to identifying the first set of users of the online social-networking system as belonging to the first population segment and the second set of users of the online social-networking system as belonging to the second population segment, monitor account registration or login activities performed on the online social-networking system over a predetermined period of time by (1) the first set of users who are identified as belonging to the first population segment and (2) the second set of users who are identified as belonging to the second population segment that is different from the first population segment;
• store, as time series data, the monitored account registration or login activities performed on the online social-networking system over the predetermined period of time by (1) the first set of users who are identified as belonging to the first population segment and (2) the second set of users who are identified as belonging to the second population segment, wherein the account registration or login activities on the online social-networking system of two different population segments over the predetermined period of time are monitored and stored for training a machine-learning model for determining whether activities associated with a given set of users exhibit diurnal behavior pattern or non-diurnal behavior pattern;
• generate a first feature set associated with the first set of users and a second feature set associated with the second set of users by transforming the time-series data associated with each of the first and second population segments into frequency-domain data;
• using the frequency-domain data associated with each of the first and second population segments as training data to train the machine-learning model for determining whether activities associated with the given set of users exhibit diurnal behavior pattern or non-diurnal behavior pattern, the frequency-domain data for training the machine-learning model comprising (1) the first feature set associated with the first population segment that exhibits diurnal behavior pattern reflecting account registration or login activities performed on the social-networking system daily and during a first time schedule and (2) the second feature set associated with the second population segment that exhibits non-diurnal behavior pattern reflecting account registration or login activities performed on the social-networking system during a second time schedule that is different from the first time schedule of the first population segment;
• and
• determine, using the trained machine-learning model, whether activities performed by a third set of users on the online social-networking system exhibit diurnal behavior pattern or non-diurnal behavior pattern.

A system comprising:

• one or more processors;
• and a memory coupled to the processors comprising instructions executable by the processors, the processors being operable when executing the instructions to:
• identify, based on predetermined criteria, a first set of users of an online social-networking system as belonging to a first population segment and a second set of users of the online social-networking system as belonging to a second population segment;
• in response to identifying the first set of users of the online social-networking system as belonging to the first population segment and the second set of users of the online social-networking system as belonging to the second population segment, monitor account registration or login activities performed on the online social-networking system over a predetermined period of time by (1) the first set of users who are identified as belonging to the first population segment and (2) the second set of users who are identified as belonging to the second population segment that is different from the first population segment;
• store, as time series data, the monitored account registration or login activities performed on the online social-networking system over the predetermined period of time by (1) the first set of users who are identified as belonging to the first population segment and (2) the second set of users who are identified as belonging to the second population segment, wherein the account registration or login activities on the online social-networking system of two different population segments over the predetermined period of time are monitored and stored for training a machine-learning model for determining whether activities associated with a given set of users exhibit diurnal behavior pattern or non-diurnal behavior pattern;
• generate a first feature set associated with the first set of users and a second feature set associated with the second set of users by transforming the time-series data associated with each of the first and second population segments into frequency-domain data;
• using the frequency-domain data associated with each of the first and second population segments as training data to train the machine-learning model for determining whether activities associated with the given set of users exhibit diurnal behavior pattern or non-diurnal behavior pattern, the frequency-domain data for training the machine-learning model comprising (1) the first feature set associated with the first population segment that exhibits diurnal behavior pattern reflecting account registration or login activities performed on the social-networking system daily and during a first time schedule and (2) the second feature set associated with the second population segment that exhibits non-diurnal behavior pattern reflecting account registration or login activities performed on the social-networking system during a second time schedule that is different from the first time schedule of the first population segment;
• and
• determine, using the trained machine-learning model, whether activities performed by a third set of users on the online social-networking system exhibit diurnal behavior pattern or non-diurnal behavior pattern.