Frequency-based reputation
View Patent ↗In an example embodiment, there is disclosed an anti-malware system and method that analyzes applications and other executable objects at the subroutine level. Each subroutine may be assigned an execution frequency score, which may be based on simulated execution in an insulated environment, on real-world operating conditions, or on static analysis. Each subroutine may be assigned an execution frequency score based on how frequently it is executed. Based on this score, each subroutine may also be assigned a reputation score. To aid in cross-referencing to occurrences of the same subroutine in other applications, the subroutine may also be assigned a pseudo-unique identifier, such as a fuzzy fingerprint.
1. An apparatus for performing frequency-based classification of executable objects, comprising:
a hardware platform comprising a processor communicatively coupled to a memory;
a network interface; and
a reputation client engine to operate on the hardware platform, and configured to:
parse an executable object into a plurality of subroutines;
assign a pseudo-unique identifier to a subroutine;
find subroutines with the same pseudo-unique identifier having previously-calculated execution frequencies;
aggregate the execution frequencies of the subroutines;
assign an execution frequency score to each subroutine according to the aggregated execution frequencies;
provide the execution frequency scores to a service via the network interface; and
receive via the network interface a reputation score for the executable object based at least in part on the execution frequency scores of the subroutines of the executable object.
2. The apparatus of claim 1 , wherein the assigning an execution frequency score to the subroutines further comprises running the executable object and tallying calls to the subroutines.
3. The apparatus of claim 1 , wherein the reputation client engine is further operable to run the executable object in an insulated environment.
4. The apparatus of claim 1 , wherein the parsing the executable object into a plurality of subroutines further comprises modifying a copy of the executable object to inject frequency counters into the subroutines.
5. A reputation server comprising:
a hardware platform comprising a processor communicatively coupled to a memory;
a network interface; and
a reputation server engine to operate on the hardware platform and configured to:
parse an executable object into a plurality of subroutines;
assign a pseudo-unique identifier to a subroutine;
find subroutines with the same pseudo-unique identifier having previously-calculated execution frequencies;
aggregate the execution frequencies of the subroutines;
determine an execution frequency for each subroutine according to the aggregated execution frequencies;
assign a reputation score to each subroutine based at least in part on the execution frequency for the subroutine; and
assign the executable object a reputation score based at least in part on the reputations of the subroutines.
6. The reputation server of claim 5 , wherein the reputation server engine is further operable to:
compute a reputation score for the executable object based at least in part on the reputation scores for the plurality of subroutines; and
distribute at least one of the reputation scores to a plurality of reputation clients via the network interface.
7. The reputation server of claim 6 , wherein assigning a reputation score to each subroutine further comprises assigning an execution frequency score to each subroutine.
8. The reputation server of claim 6 , wherein parsing the executable object into a plurality of subroutines further comprises receiving from a reputation client data concerning the plurality of subroutines.
9. The reputation server of claim 6 , wherein assigning a reputation score to each subroutine further comprises receiving an execution frequency score for each subroutine from a reputation client.
10. The reputation server of claim 6 , wherein assigning a reputation score to each subroutine further comprises:
assigning each subroutine a pseudo-unique fingerprint; and
using the pseudo-unique fingerprint to query a reputation database for a frequency or reputation score for a matching subroutine.
11. The reputation server of claim 6 , wherein assigning a reputation score to each subroutine further comprises receiving an execution frequency score for each subroutine, the execution frequency score comprising execution frequency across a plurality of executable objects.
12. The reputation server of claim 6 , wherein the reputation server engine is further operable to:
run the executable object inside an insulated environment; and
assign subroutines frequency scores based on their execution frequencies.
13. The reputation server of claim 6 , wherein assigning a reputation score to each subroutine further comprises statically analyzing the executable object.
14. The reputation server of claim 6 , wherein the reputation server engine is further operable to flag executable objects with low reputation scores for additional deep analysis.
15. The reputation server of claim 6 , wherein the reputation for the executable object is based at least in part on the reputations of the plurality of subroutines.
16. The reputation server of claim 6 , wherein the reputation server engine is further operable to:
determine that the executable object has a high reputation score; and
assign the plurality of subroutines a high reputation score.
17. The reputation server of claim 6 , wherein the reputation server engine is further operable to:
assign a pseudo-unique fingerprint to at least one subroutine;
compute an execution frequency score for the at least one subroutine across a plurality of executable objects; and
correlate the reputation score to the execution frequency score.
18. The reputation server of claim 17 , wherein assigning a pseudo-unique fingerprint to the at least one subroutine comprises computing a fuzzy fingerprint.
19. The reputation server of claim 6 , wherein the reputation server engine is further operable to identify subroutines with an execution frequency score of zero as security risks.
20. The reputation server of claim 6 , wherein the reputation server engine is further operable to identify subroutines with low execution frequency scores as suspect.
21. A method for computing a reputation of an executable object, the method executable on a computing device and comprising:
parsing an executable object into a plurality of subroutines;
assigning a pseudo-unique identifier to a subroutine;
finding subroutines with the same pseudo-unique identifier having previously-calculated execution frequencies;
aggregating the execution frequencies of the subroutines;
assigning to each subroutine an execution frequency score according to the aggregated execution frequencies;
assigning to the subroutines a reputation score based at least in part on the execution frequency score; and
assigning the executable object a reputation score based at least in part on the reputation scores of the subroutines.