To run this code:
- Execute "experiment_schedule.m" file. This code provides the attack to Linear SVM, RBF kernel SVM and MLP. You are also free to train your own model and apply Bayesian optimization method to attack the underlying classifier.
- To switch from SVM to MLP, simply call the function run_experiment2_diff_C, Linear SVM and Kernel SVM can be switched inside function run_experiment1_diff_C.
- A rough structure of the code is as follows:
BayesianOptimization.mis the main part where Bayesian Optimization method is implemented.
You should expect to obtain the plot of query number with different feature modification cost C.
You will need Matlab to run the whole code. Training data for the trained model in the paper is preprocessed and is ready in the folder "/Data". To train your own model on your own data, you will need to place the processed data into the "/Data" folder.