Buffer dtype mismatch about riskindroid HOT 6 CLOSED

Hi, can you provide more details? What version of the tool are you using (docker, directly on your pc - windows/ubuntu/other)? Does it work if you use the online version of the tool? If the error is for a specific apk, please provide it so I can do some debugging.

From a quick search it looks like the error may be caused by using a pc with a different architecture (scikit-learn/scikit-learn#7891, https://stackoverflow.com/questions/21033038/scikits-learn-randomforrest-trained-on-64bit-python-wont-open-on-32bit-python), but I don't know if that's your case.

from riskindroid.

I am using it directly on my pc-windows 64-bit and Python version 3.6.0

`* Serving Flask app "app" (lazy loading)

• Environment: production
  WARNING: Do not use the development server in a production environment.
  Use a production WSGI server instead.
• Debug mode: off
• Running on http://127.0.0.1:5000/ (Press CTRL+C to quit)
  127.0.0.1 - - [19/Jun/2018 19:56:37] "GET / HTTP/1.1" 200 -
  127.0.0.1 - - [19/Jun/2018 19:56:37] "GET /static/site.css HTTP/1.1" 200 -
  127.0.0.1 - - [19/Jun/2018 19:56:37] "GET /static/filestyle.min.js HTTP/1.1" 200 -
  127.0.0.1 - - [19/Jun/2018 19:56:38] "GET /favicon.ico HTTP/1.1" 404 -
  127.0.0.1 - - [19/Jun/2018 19:56:38] "GET /favicon.ico HTTP/1.1" 404 -
  C:\Users\TrickerKaim\AppData\Local\Programs\Python\Python36-32\lib\site-packages\sklearn\base.py:311: UserWarning: Trying to unpickle estimator SVC from version 0.18.1 when using version 0.19.1. This might lead to breaking code or invalid results. Use at your own risk.
  UserWarning)
  C:\Users\TrickerKaim\AppData\Local\Programs\Python\Python36-32\lib\site-packages\sklearn\base.py:311: UserWarning: Trying to unpickle estimator MultinomialNB from version 0.18.1 when using version 0.19.1. This might lead to breaking code or invalid results. Use at your own risk.
  UserWarning)
  [2018-06-19 19:56:44,566] ERROR in app: Exception on /upload [POST]
  Traceback (most recent call last):
  File "C:\Users\TrickerKaim\AppData\Local\Programs\Python\Python36-32\lib\site-packages\flask\app.py", line 2292, in wsgi_app
  response = self.full_dispatch_request()
  File "C:\Users\TrickerKaim\AppData\Local\Programs\Python\Python36-32\lib\site-packages\flask\app.py", line 1815, in full_dispatch_request
  rv = self.handle_user_exception(e)
  File "C:\Users\TrickerKaim\AppData\Local\Programs\Python\Python36-32\lib\site-packages\flask\app.py", line 1718, in handle_user_exception
  reraise(exc_type, exc_value, tb)
  File "C:\Users\TrickerKaim\AppData\Local\Programs\Python\Python36-32\lib\site-packages\flask_compat.py", line 35, in reraise
  raise value
  File "C:\Users\TrickerKaim\AppData\Local\Programs\Python\Python36-32\lib\site-packages\flask\app.py", line 1813, in full_dispatch_request
  rv = self.dispatch_request()
  File "C:\Users\TrickerKaim\AppData\Local\Programs\Python\Python36-32\lib\site-packages\flask\app.py", line 1799, in dispatch_request
  return self.view_functionsrule.endpoint
  File "app/app.py", line 92, in upload_apk
  rid = RiskInDroid()
  File "C:\Users\Anyone\Documents\MNIT Project\riskdroid\RiskInDroid\app\RiskInDroid.py", line 220, in init
  self.trained_models.append(joblib.load(os.path.join(self.saved_models_dir, _model_name)))
  File "C:\Users\TrickerKaim\AppData\Local\Programs\Python\Python36-32\lib\site-packages\sklearn\externals\joblib\numpy_pickle.py", line 578, in load
  obj = _unpickle(fobj, filename, mmap_mode)
  File "C:\Users\TrickerKaim\AppData\Local\Programs\Python\Python36-32\lib\site-packages\sklearn\externals\joblib\numpy_pickle.py", line 508, in _unpickle
  obj = unpickler.load()
  File "C:\Users\TrickerKaim\AppData\Local\Programs\Python\Python36-32\lib\pickle.py", line 1050, in load
  dispatchkey[0]
  File "C:\Users\TrickerKaim\AppData\Local\Programs\Python\Python36-32\lib\site-packages\sklearn\externals\joblib\numpy_pickle.py", line 341, in load_build
  self.stack.append(array_wrapper.read(self))
  File "C:\Users\TrickerKaim\AppData\Local\Programs\Python\Python36-32\lib\site-packages\sklearn\externals\joblib\numpy_pickle.py", line 184, in read
  array = self.read_array(unpickler)
  File "C:\Users\TrickerKaim\AppData\Local\Programs\Python\Python36-32\lib\site-packages\sklearn\externals\joblib\numpy_pickle.py", line 108, in read_array
  array = pickle.load(unpickler.file_handle)
  File "sklearn\tree_tree.pyx", line 601, in sklearn.tree._tree.Tree.cinit
  ValueError: Buffer dtype mismatch, expected 'SIZE_t' but got 'long long'
  127.0.0.1 - - [19/Jun/2018 19:56:44] "POST /upload HTTP/1.1" 500 -`

Requirements Installed

Flask-SQLAlchemy==2.3.2
Flask==1.0.2
numpy==1.14.1
scikit-learn==0.19.1
scipy==1.1.0
tqdm==4.23.4
uWSGI==2.0.14

And yes online version is working fine on same apk file.
Any suggestion

from riskindroid.

You are using a different version of the scikit-learn library, 0.19.1, while the tool uses an older version, 0.18.1. You have 2 options:

• install the same versions of the libraries used by RiskInDroid (see requirements.txt file in this reporsitory)
• delete the models saved in the app/models/ directory, so that the next time you run the tool you will force the training of the models using your version of the library (this might take some minutes)

from riskindroid.

Thanks a lot now its working.
I just changed the version of scikit-learn to 0.18.1.

from riskindroid.

Can you please explain me how you have calculated Risk for the data set in RiskinDroid.py
i.e Using static impact i got it, using dynamic impact got it but how you have calculated based on classifier
RIV value of each classifier?
Is it
Risk(i) = likelihood(i) * dynamic impact(i)
or something more is added

from riskindroid.

From the classifiers contained in the scikit-learn Python library we evaluated only those which provide probability estimation. We then created a training set, composed by both malware and applications downloaded from the official Play Store. We used this training set to analyze the accuracy of the classifiers and then we chose to use only the 4 best performing ones (in terms of accuracy). Since each classifier outputs a probability, the final risk score (for each app) is obtained by averaging the results given by the 4 classifiers. We used only the classifiers that provide a predict_proba method (e.g., Support Vector Machines - http://scikit-learn.org/stable/modules/generated/sklearn.svm.SVC.html#sklearn.svm.SVC.predict_proba). To calculate the accuracy, we considered classification output above 50% as malware and below 50% as not malware, then we used the formula (correct predictions / total predictions) to find the accuracy for each classifier over the training set (using 10 fold cross validation). After choosing the 4 classifiers with the highest accuracy, we still used the predict_proba method to obtain a risk value for the applications not belonging to the training set (the risk value is the % of belonging to the malware class, taken from the output of predict_proba method), so the final risk result for each app is the average output of the predict_proba method of the 4 chosen classifiers.

from riskindroid.