embedd_rnn.py

#!/usr/bin/env python2
# -*- coding: utf-8 -*-
"""
Created on Fri May 24 00:10:01 2019

@author: uranusq
"""
import json
from keras.models import model_from_json
from keras_preprocessing.text import tokenizer_from_json
import numpy as np
from keras.preprocessing import sequence

mode = 100


with open("model_100api_600_genLegal.json", "r") as f:
    model = model_from_json(f.read())
model.load_weights('model_100api_600_check.h5')

with open("tokenizer.json", "rb") as f:
    data = json.load(f)
    tokenizer = tokenizer_from_json(data)


def predict_rnn(api_calls):
    if len(api_calls) > mode:
        api_calls = api_calls[-mode:]
    tokens = tokenizer.texts_to_sequences(api_calls)

    if len(tokens) < mode:
        for i in range(mode-len(tokens)):
            tokens.append([0])
    tokens = sequence.pad_sequences(tokens)
    tokens = np.reshape(tokens, (1, mode))
    #tokens = np.fliplr(tokens)
    return model.predict(tokens)[0]  


if __name__ == "__main__":
    api_calls = ["NtClose", "LdrLoadDll", "cryptacquirecontexta"]
    res = predict_rnn(api_calls)
    if res > 0.5:
        print("Malicious: ", res)
    else:
        print("Ok: ", res)