Probabilistic anti-encrypting malware protections for cloud-based file systems
Disclosed are systems and associated methods for protecting systems against software intended to damage or disable computers and computer systems, commonly called “malware” especially encrypting malware. Both agent-based and agentless implementations allow the identification of malware and the protection of local and cloud-based data by observing changes to filesystem structure and the information content of files, with no need to scan memory or interfere with the processing of individual processes. The data permeability of the protected system can be dynamically changed, allowing user-directed changes to be committed to storage and backed up, while adverse or potentially adverse changes are quarantined.
1. A method for detecting and mitigating attacks by malware, the method comprising:
receiving a notification of a file change event to a file system;
determining characteristics associated with the file change event, wherein determining the characteristics associated with the file change event include determining a similarity of a changed file resulting from the file change event with respect to an original file, wherein the original file is an original version of the changed file, prior to the file change event;
classifying the characteristics of the file change event to determine a malware probability estimate, wherein the malware probability estimate measures a system confidence that the file change event resulted from execution of a malware process;
diverting storage of the file change event from a protected storage to a temporary storage area in response to the malware probability estimate exceeding a threshold, wherein the temporary storage area is non-local with respect to the protected storage; and changes associated with the file change event are executed in the temporary storage area, wherein the changes associated with the file change event are propagated to the protected storage if the file change event is determined to not be malware;
generating a snapshot of the file system in response to the file change event; and
intercepting malware based on a comparison of a plurality of the snapshots.
2. The method of claim 1 , wherein the temporary storage area is physically distinct from the protected storage.
3. The method of claim 2 , wherein the temporary storage area is physically remote from the protected storage.
4. The method of claim 1 , further comprising moderating access to a first plurality of files organized in a first file system, wherein the first plurality of files are stored in the protected storage.
5. The method of claim 4 , further comprising creating a substantially identical copy of the first plurality of files with a second plurality of files stored in the temporary storage.
6. The method of claim 1 , further comprising characterizing the file change event, the characterization including at least two or more of a group selected from file creation, file deletion, file rename, file move, file type change, file entropy change, and file content change.
7. The method of claim 1 , wherein generating the snapshot includes generating a plurality of different consistent snapshots of a file system tree of the file system.
8. The method of claim 7 , further comprising detecting patterns of change based on changes between a first snapshot in the plurality of snapshots and a second snapshot in the plurality of snapshots.
9. The method of claim 1 , wherein a locality-based hashing algorithm is used to process the original file and the changed file and to place them into a similar vector space.
10. The method of claim 9 , further comprising detecting malware based on an alignment of vectors in the similar vector space.
11. The method of claim 1 , further comprising:
compressing the original file and the changed file to increase an information content of the information; and
detecting malware based on a similarity between the compressed original file and the compressed changed file.
12. The method of claim 11 , further comprising detecting malware based on a similarity difference of more than 20 percent.