PhishFOE Dataset

1. Load the Dataset

import pandas as pd

df = pd.read_csv("PhishFOE_dataset.csv")

2. Feature Enggineering Function

def SuspiciousCharRatio(data):

    data['SuspiciousCharRatio'] = (

            data['NoOfObfuscatedChar'] +

            data['NoOfEqual'] +

            data['NoOfQmark'] +

            data['NoOfAmp']

        ) / data['URLLength']

    return data

def URLComplexityScore(data):

    # Calculate the first term: (URLLength + NoOfSubDomain + NoOfObfuscatedChar) / URLLength

    first_term = (

        data['URLLength'] + 

        data['NoOfSubDomain'] + 

        data['NoOfObfuscatedChar']

    ) / data['URLLength']   

    # Calculate the second term: (NoOfEqual + NoOfAmp) / (NoOfQmark + 1)

    second_term = (

        data['NoOfEqual'] + 

        data['NoOfAmp']

    ) / (data['NoOfQmark'] + 1)

    data['URLComplexityScore'] = first_term + second_term

    return data

def HTMLContentDensity(data):

    data['HTMLContentDensity'] = (

            data['LineLength'] + data['NoOfImage']

        ) / (

            data['NoOfJS'] + data['NoOfCSS'] + data['NoOfiFrame'] + 1

        )    

    return data

def InteractiveElementDensity(data):

    data['InteractiveElementDensity'] = (

            data['HasSubmitButton'] +

            data['HasPasswordField'] +

            data['NoOfPopup']

        ) / (

            data['LineLength'] + data['NoOfImage']

        )

    return data

3. Label Encoder

d = defaultdict(LabelEncoder)

df = df.apply(lambda x: d[x.name].fit_transform(x))

df_FS = df.copy()

y = df_FS['label']

X = df_FS.drop(columns=['label','URL'])

4. Boruta Feature Selector

def ranking(ranks, names, order=1):

    minmax = MinMaxScaler()

    ranks = minmax.fit_transform(order * np.array([ranks]).T).T[0]

    ranks = map(lambda x: round(x, 2), ranks)

    return dict(zip(names, ranks))

# Initialize RandomForestClassifier for Boruta

rf = RandomForestClassifier(

    n_jobs=-1, 

    class_weight="balanced_subsample", 

    max_depth=5, 

    n_estimators="auto"  # Adjust n_estimators to a specific value since 'auto' is not valid for this parameter

)

# Initialize Boruta Feature Selector

feat_selector = BorutaPy(rf, n_estimators='auto', random_state=1)

# Fit Boruta to the dataset

feat_selector.fit(X.values, y.values.ravel())

selected_features = X.columns[feat_selector.support_].tolist()

# Use only the selected features

X_selected = X[selected_features]

5. Split for model training

X_train, X_test, y_train, y_test = train_test_split(X_selected, y, test_size = 0.2, random_state = 42)

X_train.shape, y_train.shape, X_test.shape, y_test.shape