Detecting computer security risk based on previously observed communications

1. A method of detecting security risk, comprising:

receiving information about an electronic message that is from a sender for an intended recipient;

determining whether the sender of the electronic message has an established relationship with the intended recipient, wherein the established relationship is based at least in part on previous email messages between the sender and the intended recipient;

in response to the determination that the sender of the electronic message has the established relationship with the intended recipient, using a processor to analyze the electronic message based at least in part on previously observed communications between the sender and the intended recipient to determine a security risk of the electronic message for the intended recipient using a statistically analyzed result identifying a likelihood of existence of a cluster of two or more email header items by tracking and matching combinations of corresponding email header items from email headers of previous email messages sent from the sender, wherein the cluster of the combination of two or more of the email header items include a mail user agent (MUA) metadata item, a time zone, an IP address, X-header metadata information, or an identification of a supported character set; and

based on the determined security risk of the electronic message, performing a security action, if applicable.

2. The method of claim 1 , wherein determining whether the sender of the electronic message has the established relationship with the intended recipient includes determining whether the sender has previously sent the intended recipient a threshold number of messages.

3. The method of claim 1 , wherein determining whether the sender of the electronic message has the established relationship with the intended recipient includes determining whether the sender had been in communication with the intended recipient for at least a threshold amount of time.

4. The method of claim 1 , wherein determining whether the sender of the electronic message has the established relationship with the intended recipient includes obtaining information about the sender or user relationships of the intended recipient from a service that tracks social network relationship data.

5. The method of claim 1 , wherein determining whether the sender of the electronic message has the established relationship with the intended recipient includes determining a trust score associated with the sender with respect to the intended recipient of the message.

6. The method of claim 1 , wherein analyzing the electronic message based at least in part on the previously observed communications between the sender and the intended recipient to determine the security risk of the electronic message for the intended recipient includes using profiles of body content included in previous messages received from the sender.

7. The method of claim 1 , wherein analyzing the electronic message based at least in part on the previously observed communications between the sender and the intended recipient to determine the security risk of the electronic message for the intended recipient includes using profiles of header content included in the previous messages received from the sender.

8. The method of claim 1 , wherein analyzing the electronic message based at least in part on the previously observed communications between the sender and the intended recipient to determine the security risk of the electronic message for the intended recipient includes determining a plurality of risk component scores and combining the plurality of risk component scores to determine an overall risk score of the security risk of the electronic message for the intended recipient.

9. The method of claim 1 , wherein analyzing the electronic message based at least in part on the previously observed communications between the sender and the intended recipient to determine the security risk of the electronic message for the intended recipient includes determining a risk component score based on whether the message includes a keyword and whether the keyword or a related keyword was included in a previous message from the sender.

10. The method of claim 1 , wherein determining the security risk of the electronic message for the intended recipient includes computing a Bayesian probability score of a particular feature of message being consistent with past observations of previous messages from the sender.

11. The method of claim 1 , wherein analyzing the electronic message based at least in part on the previously observed communications between the sender and the intended recipient to determine the security risk of the electronic message for the intended recipient includes determining a probability a plurality of aspects of the electronic message have been exhibited together in previous messages from the sender for the intended recipient.

12. The method of claim 1 , wherein performing the security action includes selecting based on the determined security risk of the electronic message, which security action option to perform among a plurality of security action options.

13. The method of claim 1 , wherein performing the security action includes determining that a risk score associated with the security risk of the electronic message is below a threshold value and allowing the intended recipient to fully access the electronic message by delivering the electronic message to the intended recipient.

14. The method of claim 1 , wherein performing the security action includes determining that a risk score associated with the security risk of the electronic message is above a threshold value and modifying the electronic message prior to allowing the intended recipient to access a modified version of the electronic message.

15. The method of claim 1 , wherein performing the security action includes performing one or more of the following: sending a verification challenge to an alternative contact of the sender; performing additional analysis of the electronic message; quarantining the electronic message; blocking the electronic message; executing an executable included in the electronic message in a sandbox or a virtual machine; adding a warning to the electronic message; and moving the electronic message to a different folder.

16. A system of detecting security risk, comprising:

a hardware processor configured to:

receive information about an electronic message that is from a sender for an intended recipient;

determine whether the sender of the electronic message has an established relationship with the intended recipient, wherein the established relationship is based at least in part on previous email messages between the sender and the intended recipient;

in response to the determination that the sender of the electronic message has the established relationship with the intended recipient, analyze the electronic message based at least in part on previously observed communications between the sender and the intended recipient to determine a security risk of the electronic message for the intended recipient using a statistically analyzed result identifying a likelihood of existence of a cluster of two or more email header items by tracking and matching combinations of corresponding email header items from email headers of previous email messages sent from the sender, wherein the cluster of the combination of two or more of the email header items include a mail user agent (MUA) metadata item, a time zone, an IP address, X-header metadata information, or an identification of a supported character set; and

based on the determined security risk of the electronic message, perform a security action, if applicable; and

a memory coupled to the processor and configured to provide the processor with instructions.

17. The system of claim 16 , wherein analyzing the electronic message based at least in part on the previously observed communications between the sender and the intended recipient to determine the security risk of the electronic message for the intended recipient includes using profiles of body content included in previous messages received from the sender.

18. The system of claim 16 , wherein analyzing the electronic message based at least in part on the previously observed communications between the sender and the intended recipient to determine the security risk of the electronic message for the intended recipient includes using profiles of header content included in the previous messages received from the sender.

19. The system of claim 16 , wherein analyzing the electronic message based at least in part on the previously observed communications between the sender and the intended recipient to determine the security risk of the electronic message for the intended recipient includes determining a plurality of risk component scores and combining the plurality of risk component scores to determine an overall risk score of the security risk of the electronic message for the intended recipient.

20. A computer program product, the computer program product being embodied in a non-transitory computer readable storage medium and comprising computer instructions executable by a hardware processor for detecting security risk based on by performing:

receiving information about an electronic message that is from a sender for an intended recipient;

determining whether the sender of the electronic message has an established relationship with the intended recipient, wherein the established relationship is based at least in part on previous email messages between the sender and the intended recipient;

in response to the determination that the sender of the electronic message has the established relationship with the intended recipient, using a processor to analyze the electronic message based at least in part on previously observed communications between the sender and the intended recipient to determine a security risk of the electronic message for the intended recipient using a statically analyzed result identifying a likelihood of existence of a cluster of two or more email header items by tracking and matching combinations of corresponding email header items from email headers of previous email messages sent from the sender, wherein the cluster of the combination of two or more of the email header items include a mail user agent (MUA) metadata item, a time zone, an IP address, X-header metadata information, or an identification of a supported character set; and

based on the determined security risk of the electronic message, performing a security action, if applicable.