Monitoring user status by comparing public and private activities
View Patent ↗One embodiment of the present invention provides a system for detecting anomalous correlations between public and private activities of a user. During operation, the system collects public and private activity data associated with the user. The system generates a series of feature pairs, each feature pair including a public feature vector and a private feature vector generated from the activity data. Each respective feature pair corresponds to a respective point in time. The system generates a model to determine whether there is an anomaly in a correlation between the user's public and private activity data. The model is associated with a normal correlation between the user's public and private activity data over a period of time. The system collects additional public and private activity data and applies the model to determine whether there is an anomaly. The system may issue an alert in response to detecting an anomaly.
1. A computer-executable method for detecting anomalous correlations between public and private activities of a user, comprising:
collecting public activity data and private activity data associated with the user;
generating a series of feature pairs, each feature pair including a public feature vector generated from public activity data and a private feature vector generated from private activity data, wherein a respective feature pair corresponds to a respective point in time;
generating a model, based on at least the series of feature pairs, to determine whether there is an anomaly in a correlation between the user's public activity data and the user's private activity data, wherein the model is associated with a correlation between the user's public activity data and the user's private activity data over a period of time associated with the series of feature pairs;
collecting additional public activity data and additional private activity data associated with the user;
applying the model to determine whether there is an anomaly in the correlation between the additional public activity data and the additional private activity data; and
issuing an alert in response to detecting an anomaly in the correlation between the additional public activity data and the additional private activity data of the user.
2. The method of claim 1 , wherein the model is a one-class support vector machine; and
wherein generating the model comprises training a one-class support vector machine with the series of feature pairs.
3. The method of claim 1 , wherein the model is a rules-based model, further comprising:
obtaining rules for detecting anomalies in the public activity data and the private activity data; and
wherein the generating the model comprises generating the rules-based model using the obtained rules and the series of feature pairs.
4. The method of claim 1 , further comprising generating the public feature vectors and the private feature vectors with features that include one or more of:
count of punctuation characters for each predetermined number of words in the text;
count of average number of part-of-speech (POS) tags;
count of individual persons that the user sent emails to;
count of social networking friends that the user left a comment for or showed support or enjoyment of;
count of received and/or sent emails;
count of shared posts, photos, and videos;
count of newly emailed persons;
count of recently added friends on a social networking service;
topic distribution of written content including emails and discussions on the social networking service; and
changing trends of one or more features.
5. The method of claim 1 , wherein detecting an anomaly includes detecting a change beyond a predetermined threshold in one or more of the following correlations over a predetermined period of time:
correlation of private activity sentiment level and public activity sentiment level;
correlation between a count of events on a social networking service and a count of sent emails;
correlation between a number of new email communication partners and a number of new friends added to the social networking service;
correlation between a rate of adding new connections on the social networking service and a rate of adding new email communication partners;
correlation between frequency of use of the social networking service and frequency of use of private communications;
correlation between a number of emails and a number of posts on the social networking service;
correlation between topics that the user mentions in private communications and topics that the user mentions on the social networking service; and
correlation between use of language on the social networking service and use of language in private communications.
6. The method of claim 5 , further comprising:
calculating a feature with a sliding window by analyzing data within a given range before and after each point in time.
7. A non-transitory computer-readable storage medium storing instructions that when executed by a computer cause the computer to perform a method for detecting anomalous correlations between public and private activities of a user, the method comprising:
collecting public activity data and private activity data associated with the user;
generating a series of feature pairs, each feature pair including a public feature vector generated from public activity data and a private feature vector generated from private activity data, wherein a respective feature pair corresponds to a respective point in time;
generating a model, based on at least the series of feature pairs, to determine whether there is an anomaly in a correlation between the user's public activity data and the user's private activity data, wherein the model is associated with a correlation between the user's public activity data and the user's private activity data over a period of time associated with the series of feature pairs;
collecting additional public activity data and additional private activity data associated with the user;
applying the model to determine whether there is an anomaly in the correlation between the additional public activity data and the additional private activity data; and
issuing an alert in response to detecting an anomaly in the correlation between the additional public activity data and the additional private activity data of the user.
8. The non-transitory computer-readable storage medium of claim 7 , wherein the model is a one-class support vector machine; and
wherein generating the model comprises training a one-class support vector machine with the series of feature pairs.
9. The non-transitory computer-readable storage medium of claim 7 , wherein the model is a rules-based model, and the method further comprises:
obtaining rules for detecting anomalies in the public activity data and the private activity data; and
wherein the generating the model comprises generating the rules-based model using the obtained rules and the series of feature pairs.
10. The non-transitory computer-readable storage medium of claim 7 , wherein the method further comprises generating the public feature vectors and the private feature vectors with features that include one or more of:
count of punctuation characters for each predetermined number of words in the text;
count of average number of POS tags;
count of individual persons that the user sent emails to;
count of social networking friends that the user left a comment for or showed support or enjoyment of;
count of received and/or sent emails;
count of shared posts, photos, and videos;
count of newly emailed persons;
count of recently added friends on a social networking service;
topic distribution of written content including emails and discussions on the social networking service; and
changing trends of one or more features.
11. The non-transitory computer-readable storage medium of claim 7 , wherein detecting an anomaly includes detecting a change beyond a predetermined threshold in one or more of the following correlations over a predetermined period of time:
correlation of private activity sentiment level and public activity sentiment level;
correlation between a count of events on a social networking service and a count of sent emails;
correlation between a number of new email communication partners and a number of new friends added to the social networking service;
correlation between a rate of adding new connections on the social networking service and a rate of adding new email communication partners;
correlation between frequency of use of the social networking service and frequency of use of private communications;
correlation between a number of emails and a number of posts on the social networking service;
correlation between topics that the user mentions in private communications and topics that the user mentions on the social networking service; and
correlation between use of language on the social networking service and use of language in private communications.
12. The non-transitory computer-readable storage medium of claim 7 , wherein the method further comprises:
calculating a feature with a sliding window by analyzing data within a given range before and after each point in time.
13. A computing system for detecting anomalous correlations between public and private activities of a user, the system comprising:
one or more processors,
a non-transitory computer-readable medium coupled to the one or more processors having instructions stored thereon that, when executed by the one or more processors, cause the one or more processors to perform operations comprising:
collecting public activity data and private activity data associated with the user;
generating a series of feature pairs, each feature pair including a public feature vector generated from public activity data and a private feature vector generated from private activity data, wherein a respective feature pair corresponds to a respective point in time;
generating a model, based on at least the series of feature pairs, to determine whether there is an anomaly in a correlation between the user's public activity data and the user's private activity data, wherein the model is associated with a correlation between the user's public activity data and the user's private activity data over a period of time associated with the series of feature pairs;
collecting additional public activity data and additional private activity data associated with the user;
applying the model to determine whether there is an anomaly in the correlation between the additional public activity data and the additional private activity data; and
issuing an alert in response to detecting an anomaly in the correlation between the additional public activity data and the additional private activity data of the user.
14. The computing system of claim 13 , wherein the model is a one-class support vector machine; and
wherein generating the model comprises training a one-class support vector machine with the series of feature pairs.
15. The computing system of claim 13 , wherein the model is a rules-based model, wherein the non-transitory computer-readable storage medium stores additional instructions that, when executed, cause the one or more processors to perform additional steps comprising:
obtaining rules for detecting anomalies in the public activity data and the private activity data; and
wherein the generating the model comprises generating the rules-based model using the obtained rules and the series of feature pairs.
16. The computing system of claim 13 , wherein the non-transitory computer-readable storage medium stores additional instructions that, when executed, cause the one or more processors to perform additional steps comprising:
generating the public feature vectors and the private feature vectors with features that include one or more of:
count of punctuation characters for each predetermined number of words in the text;
count of average number of POS tags;
count of individual persons that the user sent emails to;
count of social networking friends that the user left a comment for or showed support or enjoyment of;
count of received and/or sent emails;
count of shared posts, photos, and videos;
count of newly emailed persons;
count of recently added friends on a social networking service;
topic distribution of written content including emails and discussions on the social networking service; and
changing trends of one or more features.
17. The computing system claim 13 , wherein detecting an anomaly includes detecting a change beyond a predetermined threshold in one or more of the following correlations over a predetermined period of time:
correlation of private activity sentiment level and public activity sentiment level;
correlation between a count of events on a social networking service and a count of sent emails;
correlation between a number of new email communication partners and a number of new friends added to the social networking service;
correlation between a rate of adding new connections on the social networking service and a rate of adding new email communication partners;
correlation between frequency of use of the social networking service and frequency of use of private communications;
correlation between a number of emails and a number of posts on the social networking service;
correlation between topics that the user mentions in private communications and topics that the user mentions on the social networking service; and
correlation between use of language on the social networking service and use of language in private communications.
18. The computing system of claim 13 , wherein the non-transitory computer-readable storage medium stores additional instructions that, when executed, cause the one or more processors to perform additional steps comprising:
calculating a feature with a sliding window by analyzing data within a given range before and after each point in time.