ML for Security
=

https://ppt.cc/f6zq5x
https://drive.google.com/drive/folders/1J7WuuxGKE0oVle5WU7cnXvCq9eQfkkAl?usp=sharing

## Slide

[DeepCorr](https://arxiv.org/abs/1808.07285)

主要方案：
- 跟助教要USB拿Docker Image

sudo docker load –i <filename>
sudo docker load  < <filename> 

備用方案：
- Image:
https://drive.google.com/file/d/16HIzBegZEW7uNJrhsqTRQsWJxjC-CqSw/view?usp=sharing

Pull docker mlsec image

docker pull bletchley/mlsec:taiwanno1

下載好Ubuntu，並裝好docker後，執行(約6G大小)

docker run -p 8888:8888 -it bletchley/mlsec:taiwanno1 bash

下載完後，進到 docker 輸入

jupyter notebook --allow-root

帶有 token 的網址丟進瀏覽器，網址的部分要修改一下
將 http://(<container_id> or 127.0.0.1):8888 => http://127.0.0.1:8888

- [決策樹介紹網址](http://www.r2d3.us/visual-intro-to-machine-learning-part-1/)
- 如何切: Entropy、Gini

Payment
==
One hot encoding

df = pd.get_dummies(df, columns = ["paymentMethod"])

df.sample(3)

training/testing set

X_train, X_test, y_train, y_test = train_test_split(
    df.drop('label', axis=1), df['label'],
    test_size=0.33, random_state=17)

Model

# Initialize and train classifier model
clf = LogisticRegression().fit(X_train, y_train)

# Make predictions on test set
y_pred = clf.predict(X_test)

Docker no space left problem
https://colobu.com/2018/10/22/no-space-left-on-device-for-docker/

ImportError: cannot import name 'cross_validation'

# replace 'from sklearn import cross_validation' with:
from sklearn.model_selection import train_test_split 

algorithms = {
        "DecisionTree": tree.DecisionTreeClassifier(max_depth=10),
        "RandomForest": ensemble.RandomForestClassifier(n_estimators=50), 
        "GradientBoosting": ensemble.GradientBoostingClassifier(n_estimators=50),
        "AdaBoost": ensemble.AdaBoostClassifier(n_estimators=100),
        "GNB": naive_bayes.GaussianNB(),
    }

results = {}  
y_preds = {}  
print("Now testing algorithms\\n")  
for algo in algorithms:  
    clf = algorithms[algo]  
    clf.fit(X_train, y_train)  
    y_pred = clf.predict(X_test)  
    score = clf.score(X_test, y_test)  
    print("%s : %f %%" % (algo, score*100))  
    results[algo] = score  
    y_preds[algo] = y_pred

## Missing Value

num_orig_rows = len(df)
num_full_rows = len(df.dropna())

(num_orig_rows - num_full_rows)/float(num_orig_rows)

Drop Rows

df_droprows = df.dropna()
build_model(df_droprows)

Drop Columns

df_dropcols = df[['MonthlyIncome','Overtime','Label']]
build_model(df_dropcols)

Fill NA as -1

# use df.fillna to fill column with -1

df_sentinel = df.fillna(value=-1)
build_model(df_sentinel)

Use median

# use median instead mean
imp = Imputer(missing_values='NaN', strategy='median', axis=0)
df_imputed = pd.DataFrame(imp.fit_transform(df),
                          columns=['TotalWorkingYears', 'MonthlyIncome',
                                   'OverTime', 'DailyRate', 'Label'])
build_model(df_imputed)

## Network Moninor

from sklearn.model_selection import train_test_split

# train_test_split
x_train, x_test, y_train, y_test = train_test_split( \
    data, label, test_size=0.20, random_state=123)

# Random Forest Classifier
from sklearn import ensemble
clf = ensemble.RandomForestClassifier()
clf.fit(x_train, y_train)
pred = clf.predict(x_test)

#demostrate confusion_matrix
from sklearn.metrics import confusion_matrix
cm = confusion_matrix(y_test, pred)
print(cm)

#show accuracy_score
from sklearn import metrics
a_score = metrics.accuracy_score(y_test, pred)
print("Accuracy score:\t %.2f%%" % (a_score*100))

## nsl-kdd-classification

## sklearn-gridsearch

---
# Z3 CTF
https://gist.github.com/ekse/baee0cabbe12861443a5#file-harder_serial-py

> 喔喔等等會講 z3 喔
> https://pastebin.com/iRX8AEwc (?
> 其實沒有講，簡單帶過而已，如果有時間再說Z3原理