Techniques for scam detection and prevention
Techniques for scam detection and prevention are described. In one embodiment, an apparatus may comprise an interaction processing component operative to generate a scam message example repository; submit the scam message example repository to a natural-language machine learning component; and receive a scam message model from the natural-language machine learning component in response to submitting the scam message example repository; an interaction monitoring component operative to monitor a plurality of messaging interactions with a messaging system based on the scam message model; and determine a suspected scam messaging interaction of the plurality of messaging interactions; and a scam action component operative to perform a suspected scam messaging action with the messaging system in response to determining the suspected scam messaging interaction. Other embodiments are described and claimed.
1. A computer-implemented method, comprising:
receiving a plurality of monitored messaging interactions associated with a plurality of users including a target user, wherein at least one of the messaging interactions comprises a message associated with the target user;
determining a message reuse measure for the at least one messaging interaction, the message reuse measure to compare the reuse of the message by the target user to the reuse of the message in the plurality of monitored messaging interactions;
flagging the messaging interaction as being suspicious based on the message reuse measure;
using the flagged messaging interaction to generate a scam message model; and
determining a level of scam behavior for a new message having a message sender and a message recipient based on a degree of recognition of scam behavior in the new message according to the scam message model.
2. The computer-implemented method of claim 1 , further comprising:
monitoring real-time or near real-time messaging interactions using the scam message model.
3. The computer-implemented method of claim 1 , further comprising:
determining a low-confidence match to scam according to the new message's level of scam behavior being below a threshold, and
not taking preventative action in response to the determination of the low-confidence match.
4. The computer-implemented method of claim 1 , further comprising:
determining a high-confidence match to scam according to the new message's level of scam behavior being above a threshold, and
taking immediate preventative action in response to the determination of the high-confidence match, wherein the preventative action is determined according to the new message's level of scam behavior and the preventative action comprises at least one of:
enacting an explicit ban of the message sender,
enabling a shadow ban of the message sender,
providing scam education for the message recipient, and
promotion of scam-related tools to the message recipient.
5. The computer-implemented method of claim 1 , further comprising:
identifying a low-confidence threat according to the new message's level of scam behavior exceeding a low-confidence threshold while not exceeding a high-confidence threshold, and
queueing the low-confidence threat for review by a human administrator.
6. An apparatus, comprising:
a message reuse monitoring component operative to:
receive a plurality of monitored messaging interactions associated with a plurality of users comprising a target user, wherein at least one of the messaging interactions comprises a message associated with the target user;
a message reuse measure for the at least one messaging interaction, the message reuse measure to compare the reuse of the message by the target user to the reuse of the message in the plurality of monitored messaging interactions; and
flag the messaging interaction as being suspicious based on the message reuse measure;
a natural-language machine learning component operative to use the flagged messaging interaction to generate a scam message model; and
an interaction monitoring component operative to determine a level of scam behavior for a new message associated with a message sender and a message recipient based on a degree of recognition of scam behavior in the new message according to the scam message model.
7. The apparatus of claim 6 , further comprising:
an interaction monitoring component operative to monitor a real-time or near real time messaging interaction using the scam message model.
8. The apparatus of claim 6 , further comprising:
the interaction monitoring component operative to determine a low-confidence match to scam according to the new message's level of scam behavior being below a threshold, and
a scam action component operative to withhold from taking preventative action in response to the determination of the low-confidence match.
9. The apparatus of claim 6 , further comprising:
the interaction monitoring component operative to determine a high-confidence match to scam according to the new message's level of scam behavior being above a threshold, and
a scam action component operative to take immediate preventative action in response to the determination of the high-confidence match, wherein the preventative action is determined according to the new message's level of scam behavior and the preventative action comprises at least one of:
an explicit ban of the message sender,
enabling a shadow ban of the message sender,
providing scam education for the message recipient, and
promotion of scam-related tools to the message recipient.
10. The apparatus of claim 6 , further comprising:
a scam action component operative to identify a low-confidence threat according to the new message's level of scam behavior exceeding a low-confidence threshold while not exceeding a high-confidence threshold and queue the low-confidence threat for review by a human administrator.
11. At least one non-transitory computer-readable storage medium comprising instructions configured to, when executed, cause a system to:
receive a plurality of monitored messaging interactions associated with a plurality of users including a target user, wherein at least one of the messaging interactions comprises a message associated with the target user;
determine a message reuse measure for the at least one messaging interaction, the message reuse measure to compare the reuse of the message by the target user to the reuse of the message in the plurality of monitored messaging interactions;
flag the messaging interaction as being suspicious based on the message reuse measure
use the flagged messaging interaction to generate a scam message model; and
determine a level of scam behavior for a new message having a message sender and a message recipient based on a degree of recognition of scam behavior in the new message according to the scam message model.
12. The non-transitory computer-readable storage medium of claim 11 , the instructions further configured to, when executed, cause the system to:
monitor real-time or near real-time messaging interactions using the scam message model.
13. The non-transitory computer-readable storage medium of claim 11 , the instructions further configured to, when executed, cause the system to:
determine a high-confidence match to scam according to the new message's level of scam behavior being above a threshold, and
take immediate preventative action in response to the determination of the high-confidence match, wherein the preventative action is determined according to the new message's level of scam behavior and the preventative action comprises at least one of:
enacting an explicit ban of the message sender,
enabling a shadow ban of the message sender,
providing scam education for the message recipient, and
promotion of scam-related tools to the message recipient.
14. The non-transitory computer-readable storage medium of claim 11 , the instructions further configured to, when executed, cause the system to:
identify a low-confidence threat according to the new message's level of scam behavior exceeding a low-confidence threshold while not exceeding a high-confidence threshold, and queue the low-confidence threat for review by a human administrator.