Methods and apparatus to perform malware detection using a generative adversarial network
Methods, apparatus, systems and articles of manufacture are disclosed to perform malware detection using a generative adversarial network. An example apparatus includes a first encoder network to encode an input sample into a first encoded sample, the first encoder network implemented using a multilayer perception (MLP) network, a generator network to reconstruct the first encoded sample to generate a reconstructed sample, a discriminator network to, in response to obtaining the first encoded sample and the reconstructed sample, generate a loss function based on the reconstructed sample and the input sample, and an optimization processor to, when the loss function satisfies a threshold loss value, classify the input sample as malicious.
1. An apparatus comprising:
memory;
instructions in the apparatus; and
processor circuitry to execute the instructions to:
encode an input sample into a first encoded sample, the input sample encoded using a first multilayer perception (MLP) network;
reconstruct the first encoded sample to generate a reconstructed sample;
encode the reconstructed sample using a second MLP network to generate a second encoded sample;
in response to obtaining the first encoded sample and the reconstructed sample, generate a first loss function and a second loss function based on the reconstructed sample and the input sample; and
when a weighted sum of the first loss function and the second loss function satisfies a threshold loss value, classify the input sample as malicious.
2. The apparatus of claim 1 , wherein the processor circuitry is to:
determine a third loss function based on the second encoded sample and the first encoded sample; and
in response to a weighted sum of the first loss function, the second loss function, and the third loss function satisfying the threshold loss value, classify the input sample as malicious.
3. The apparatus of claim 1 , wherein the weighted sum of the first loss function and the second loss function satisfies the threshold loss value when the weighted sum is greater than the threshold loss value.
4. The apparatus of claim 1 , wherein the first encoded sample is reconstructed using a first deep convolutional generative adversarial network (DCGAN), and the first loss function is generated using a second DCGAN.
5. The apparatus of claim 1 , wherein the input sample is a portable executable file.
6. At least one non-transitory computer readable medium comprising instructions which, when executed, cause at least one processor to at least:
encode an input sample into a first encoded sample using a first multilayer perception (MLP) network;
reconstruct the first encoded sample to generate a reconstructed sample;
encode the reconstructed sample using a second MLP network to generate a second encoded sample;
in response to obtaining the first encoded sample and the reconstructed sample, determine a first loss function based on the reconstructed sample and the input sample;
determine a second loss function based on the first encoded sample and the second encoded sample; and
when a weighted sum of the first loss function and the second loss function satisfies a threshold loss value, classify the input sample as malicious.
7. The at least one computer readable medium of claim 6 , wherein the weighted sum of the first loss function and the second loss function satisfies the threshold loss value when the weighted sum is greater than the threshold loss value.
8. The at least one computer readable medium of claim 6 , wherein the instructions, when executed, cause the at least one processor to generate the reconstructed sample and the first loss function using a first deep convolutional generative adversarial network (DCGAN) and a second DCGAN, respectively.
9. The at least one computer readable medium of claim 6 , wherein the input sample is a portable executable file.
10. The at least one computer readable medium of claim 6 , wherein the instructions, when executed, cause the at least one processor to:
determine a third loss function based on the reconstructed sample and the input sample; and
in response to a weighted sum of the first loss function, the second loss function, and the third loss function satisfying the threshold loss value, classify the input sample as malicious.
11. A method comprising:
encoding an input sample into a first encoded sample using a first multilayer perception (MLP) network;
reconstructing the first encoded sample to generate a reconstructed sample;
encoding the reconstructed sample using a second MLP network to obtain a second encoded sample;
in response to obtaining the first encoded sample and the reconstructed sample, generating a first loss function based on the reconstructed sample and the input sample;
generating a second loss function based on the first encoded sample and the second encoded sample; and
when a weighted sum of the first loss function and the second loss function is greater than a threshold loss value, classifying the input sample as malicious.
12. The method of claim 11 , wherein the reconstructed sample and the first loss function are generated using a first deep convolutional generative adversarial network (DCGAN) and a second DCGAN, respectively.
13. The method of claim 11 , wherein the input sample is a portable executable file.
14. The method of claim 11 , further including:
determining a third loss function based on the reconstructed sample and the input sample; and
in response to a weighted sum of the first loss function, the second loss function, and the third loss function satisfying the threshold loss value, classifying the input sample as malicious.