Apparatus and methods to classify malware with explainability with artificial intelligence models
Apparatus, systems, and methods to classify malware with explainability are disclosed. An example apparatus includes at least one memory; instructions in the apparatus; and processor circuitry. The example processor circuitry is to execute the instructions to: generate feature vectors from a first input; train a neural network model using a first portion of the feature vectors; add one or more fully connected layers to the trained neural network model to form a hybrid model; validate the hybrid model using a second portion of the feature vectors; and deploy the validated hybrid model as a malware classifier, the malware classifier to provide a malware classification with explainability in response to a second input.
1 . An apparatus comprising:
memory circuitry;
instructions; and
processor circuitry to execute the instructions to:
add a fully connected layer to an artificial intelligence (AI) model to form a hybrid model;
train the hybrid model;
remove the fully connected layer from the hybrid model to form a first malware classifier, the removal of the fully connected layer to expose outputs;
validate the malware classifier using the exposed outputs as features; and
deploy the malware classifier.
2 . The apparatus of claim 1 , wherein the malware classifier is to provide a malware classification with explainability in response to an input.
3 . The apparatus of claim 2 , wherein the deployed malware classifier is to provide explainability by indicating a portion of the input resulting in the malware classification.
4 . The apparatus of claim 3 , wherein the input includes a string feature of at least one of an executable or software code.
5 . The apparatus of claim 1 , wherein the fully connected layer is part of a multi-layer perceptron network, and wherein the processor circuitry is to add the multi-layer perceptron network to the AI model to form the hybrid model.
6 . The apparatus of claim 1 , wherein the AI model includes a neural network model.
7 . The apparatus of claim 6 , wherein the neural network model includes a convolutional neural network model with max pooling.
8 . The apparatus of claim 1 , wherein the processor circuitry is to train the hybrid model using at least one of malware identification rules or threat techniques.
9 . The apparatus of claim 1 , wherein the processor circuitry is to train the hybrid model using static features extracted from a portable executable.
10 . The apparatus of claim 1 , wherein the processor circuitry is to update the malware classifier to deploy an updated malware classifier based on at least one of feedback or new input.
11 . A non-transitory computer readable storage medium comprising instructions which, when executed, cause processor circuitry to at least:
add a fully connected layer to an artificial intelligence (AI) model to form a hybrid model;
train the hybrid model;
remove the fully connected layer from the hybrid model to form a first malware classifier, the removal of the fully connected layer to expose outputs;
validate the malware classifier using the exposed outputs as features; and
deploy the malware classifier.
12 . The non-transitory computer readable storage medium of claim 11 , wherein the malware classifier is to provide a malware classification with explainability in response to an input.
13 . The non-transitory computer readable storage medium of claim 12 , wherein the deployed malware classifier is to provide explainability by indicating a portion of the input resulting in the malware classification.
14 . The non-transitory computer readable storage medium of claim 13 , wherein the input includes a string feature of at least one of an executable or software code.
15 . The non-transitory computer readable storage medium of claim 11 , wherein the AI model includes a convolutional neural network model with max pooling, wherein the fully connected layer is part of a multi-layer perceptron network, and wherein the processor circuitry is to add the multi-layer perceptron network to the AI model to form the hybrid model.
16 . The non-transitory computer readable storage medium of claim 11 , wherein the processor circuitry is to train the hybrid model using at least one of malware identification rules or threat techniques.
17 . The non-transitory computer readable storage medium of claim 11 , wherein the processor circuitry is to train the hybrid model using static features extracted from a portable executable.
18 . The non-transitory computer readable storage medium of claim 11 , wherein the processor circuitry is to update the malware classifier to deploy an updated malware classifier based on at least one of feedback or new input.
19 . A method comprising:
adding, by executing an instruction with processor circuitry, a fully connected layer to an artificial intelligence (AI) model to form a hybrid model;
training, by executing an instruction with the processor circuitry, the hybrid model;
removing, by executing an instruction with the processor circuitry, the fully connected layer from the hybrid model to form a first malware classifier, the removal of the fully connected layer to expose outputs;
validating, by executing an instruction with the processor circuitry, the malware classifier using the exposed outputs as features; and
deploying the malware classifier.
20 . The method of claim 19 , wherein the malware classifier is to provide a malware classification with explainability in response to an input by indicating a portion of the input resulting in the malware classification.