Adaptive malware binary rewriting
An adaptive malware writing system includes a targeting engine that classifies malware candidates as a malicious candidate or a benign candidate through a surrogate model. The surrogate model assigns a weight to each byte of the malware candidates through a saliency vector. The sum of the weights render a malware classification score. An alteration engine alters a binary form of the malware candidates classified as malware by executing a functional analysis that traces application program interface calls and memory. The alteration engine alters the binary form of the malware candidates classified as malware to render a synthesized malware. The malware analysis determines if the synthesized malware is operational by comparing an image of the synthesized malware to an image of at least one of the plurality of malware candidates. A target classifier engine identifies the vulnerabilities of a targeted computer.
1. A non-transitory machine-readable medium encoded with machine-executable instructions, wherein execution of the machine-executable instructions is for:
processing a plurality of malware candidates in a binary form that disrupts, damages, or gains unauthorized access to a targeted computer;
classifying the plurality of malware candidates through a targeting engine as a malicious candidate or a benign candidate through a surrogate model;
where the surrogate model assigns a weight to each byte of the plurality of malware candidates through a saliency vector; and
where a sum of the weight assigned to each byte of the plurality of malware candidates render a malware classification score;
altering the binary form of the plurality of malware candidates classified as the malicious candidate by an alteration engine that executes a plurality of functional analysis tools that trace application program interface calls and analyze a memory;
where the alteration engine alters the binary form of the plurality of malware candidates classified as the malicious until some of the plurality of malware candidates are classified as benign rendering a synthesized malware;
determining if the synthesized malware is operational by a malware analysis engine by comparing a plurality of images of the synthesized malware to an image of one of the plurality of malware candidates; and
generating a vulnerability report by a target classifier engine that identifies a plurality of security vulnerabilities of the targeted computer.
2. The non-transitory machine-readable medium of claim 1 , where the machine-executable instructions are executed repeatedly until some of the plurality of malware candidates are classified as benign.
3. The non-transitory machine-readable medium of claim 1 , further comprising executing a dimensionality-reduction by the targeting engine that reduces a dimensionality of the plurality of malware candidates.
4. The non-transitory machine-readable medium of claim 1 , where the binary form comprises a stripped binary form.
5. The non-transitory machine-readable medium of claim 1 , where the targeting engine, the alteration engine, and the the malware analysis engine reside on a cloud.
6. The non-transitory machine-readable medium of claim 1 , where the alteration engine obfuscates the plurality of malware candidates by appending a binary form of a benign code to an end portion of each of the plurality of malware candidates.
7. The non-transitory machine-readable medium of claim 1 , where the alteration engine obfuscates the plurality of malware candidates by padding a plurality of bytes to a portion of each of the plurality of malware candidates.
8. The non-transitory machine-readable medium of claim 1 , where the alteration engine obfuscates the plurality of malware candidates by modifying a binary form of an eight-byte null-padded Unicode Transformation Format field of each of the plurality of malware candidates.
9. The non-transitory machine-readable medium of claim 1 , where the alteration engine obfuscates the plurality of malware candidates by injecting a binary form of a benign code within a mid-section of each of the plurality of malware candidates.
10. The non-transitory machine-readable medium of claim 1 , where the alteration engine obfuscates the plurality of malware candidates by appending one or more binary headers into each of the plurality of malware candidates and increasing a code count and a section count of each of the plurality of malware candidates.
11. A process facilitated by a non-transitory machine-readable medium that generates a synthesized malware, comprising:
processing a plurality of malware candidates in a binary form that disrupt, damage, and/or gain unauthorized access to a targeted computer;
classifying the plurality of malware candidates through a targeting engine as a malicious candidate or a benign candidate through a surrogate model;
where the surrogate model assigns a weight to each byte of the plurality of malware candidates through a saliency vector; and
where a sum of the weights assigned to each byte of the plurality of malware candidates render a malware classification score;
altering the binary form of the plurality of malware candidates classified as the malicious candidate by an alteration engine that executes a plurality of functional analysis tools that trace application program interface calls and analyze a memory;
where the alteration engine alters the binary form of each of the plurality of malware candidates classified as the malicious until some of the plurality of malware candidates are classified as benign rendering those plurality of malware candidates classified as benign a synthesized malware;
determining if the synthesized malware is operational by a malware analysis engine by comparing a plurality of images of the synthesized malware to one or more images of one of the plurality of malware candidates; and
generating training data from the synthesized malware that is operational and a vulnerability report by a target classifier engine; where the vulnerability report identifies a plurality of security vulnerabilities of the targeted computer.
12. An adaptive malware writing system comprising a non-transitory machine-readable medium storing machine-executable instructions, wherein execution of the machine-executable instructions causes implementation of:
a targeting engine configured to classify a plurality of malware candidates that are operational as a malicious candidate or a benign candidate through a surrogate model;
where the surrogate model assigns a weight to each byte of the plurality of malware candidates through a saliency vector; and
where a sum of the weight assigned to each byte of the plurality of malware candidates render a malware classification score;
an alteration engine configured to alter a binary form of the malware candidates classified as malicious by executing a plurality of functional analysis tools that trace application program interface calls and analyzing a memory that stores the malware candidates classified as malware;
where the alteration engine alters a binary form of the malware candidates classified as malicious until some of the plurality of malware candidates are classified as benign rendering those plurality of malware candidates classified as benign a synthesized malware;
a malware analysis engine programmed to determine if the synthesized malware is operational by comparing a plurality of images of the synthesized malware to an image of one of the plurality of malware candidates;
where the malware analysis engine discards the synthesized malware that is not operational;
a target classifier engine configured to identify a plurality of security vulnerabilities of a targeted computer and generate a training data from the synthesized malware that is operational.
13. The system of claim 12 , where the targeting engine is further programmed to execute a dimensionality-reduction by the targeting engine that reduce the dimensionality of the plurality of malware candidates.
14. The system of claim 12 , where the binary form comprises a stripped binary form.
15. The system of claim 12 , where the targeting engine, the alteration engine, and the malware analysis engine reside on a cloud.
16. The system of claim 12 , where the alteration engine obfuscates the plurality of malware candidates by appending a benign code to an end portion to the binary form of each of the plurality of malware candidates.
17. The system of claim 12 , where the alteration engine obfuscates the plurality of malware candidates by padding bytes to a portion of the binary form of each of the plurality of malware candidates.
18. The system of claim 12 , where the alteration engine obfuscates the plurality of malware candidates by modifying a binary form of an eight-byte null-padded Unicode Transformation Format field of each of the plurality of malware candidates.
19. The system of claim 12 , where the alteration engine obfuscates the plurality of malware candidates by injecting a benign code within a mid-section of the binary form of each of the plurality of malware candidates.
20. The system claim 12 where the alteration engine obfuscates the plurality of malware candidates by appending one or more binary headers into each of the plurality of malware candidates and increasing a code count and a section count of the each of the plurality of malware candidates.