app.py

import streamlit as st
import pandas as pd
import numpy as np
import pickle
import matplotlib.pyplot as plt

# Load model
with open('random_forest_classifier (1).pkl', 'rb') as file:
    model = pickle.load(file)



# Define the possible outcomes
class_labels = ['phishing', 'benign', 'defacement', 'malware']

def main():
    st.set_page_config(page_title="Malicious URL Detection", layout="wide")
    
    st.title("Malicious URL Detection")
    
    st.markdown("""
        <style>
        .main {background-color: #f5f5f5;}
        </style>
        """, unsafe_allow_html=True)
    
    st.markdown("### Enter the URL features to predict its category")
    
    # Layout: two columns for input and output
    input_col, output_col = st.columns([2, 2])

    with input_col:
        use_of_ip_address = st.number_input("Use of IP Address (0 or 1)", min_value=0, max_value=1, step=1)
        abnormal_url = st.number_input("Abnormal URL (0 or 1)", min_value=0, max_value=1, step=1)
        google_index = st.number_input("Google Index (0 or 1)", min_value=0, max_value=1, step=1)
        count_www = st.number_input("Count WWW", min_value=0)
        count_at = st.number_input("Count @", min_value=0)
        count_dir = st.number_input("Count Directory", min_value=0)
        count_embed_domian = st.number_input("Count Embedded Domain", min_value=0)
        short_url = st.number_input("Short URL (0 or 1)", min_value=0, max_value=1, step=1)
        count_https = st.number_input("Count HTTPS", min_value=0)
        count_http = st.number_input("Count HTTP", min_value=0)
        count_percent = st.number_input("Count %", min_value=0)
        count_question = st.number_input("Count ?", min_value=0)
        count_dash = st.number_input("Count -", min_value=0)
        count_equal = st.number_input("Count =", min_value=0)
        url_length = st.number_input("URL Length", min_value=0)
        hostname_length = st.number_input("Hostname Length", min_value=0)
        sus_url = st.number_input("Suspicious URL (0 or 1)", min_value=0, max_value=1, step=1)
        fd_length = st.number_input("FD Length", min_value=0)
        tld_length = st.number_input("TLD Length", min_value=0)
        count_digits = st.number_input("Count Digits", min_value=0)
        count_letters = st.number_input("Count Letters", min_value=0)

    with output_col:
        if st.button("Predict"):
            features = np.array([[use_of_ip_address, abnormal_url, google_index, count_www, count_at, 
                                  count_dir, count_embed_domian, short_url, count_https, 
                                  count_http, count_percent, count_question, count_dash, 
                                  count_equal, url_length, hostname_length, sus_url, 
                                  fd_length, tld_length, count_digits, count_letters]])
            
            prediction = model.predict(features)
            prediction_probabilities = model.predict_proba(features)[0]
            
            predicted_label = class_labels[prediction[0]]
            
            st.subheader(f"The URL is predicted to be: {predicted_label}")
            
            if predicted_label == 'phishing':
                st.error(f"The URL is predicted to be: {predicted_label}")
            elif predicted_label == 'malware':
                st.warning(f"The URL is predicted to be: {predicted_label}")
            elif predicted_label == 'defacement':
                st.warning(f"The URL is predicted to be: {predicted_label}")
            else:
                st.success(f"The URL is predicted to be: {predicted_label}")
            
            # Display prediction probabilities
            st.subheader("Prediction Probabilities:")
            prob_df = pd.DataFrame(prediction_probabilities, index=class_labels, columns=["Probability"])
            st.bar_chart(prob_df)

            # Example chart: Feature Importance
            st.subheader("Feature Importance:")
            feature_importance = model.feature_importances_
            features_names = ['use_of_ip_address', 'abnormal_url', 'google_index', 'count_www', 'count@', 
                              'count_dir', 'count_embed_domian', 'short_url', 'count_https', 
                              'count_http', 'count%', 'count?', 'count-', 'count=', 'url_length', 
                              'hostname_length', 'sus_url', 'fd_length', 'tld_length', 'count-digits', 
                              'count-letters']
            
            fig, ax = plt.subplots()
            ax.barh(features_names, feature_importance, color='skyblue')
            ax.set_xlabel('Importance')
            ax.set_title('Feature Importance')
            st.pyplot(fig)
            
            # Another example chart: Prediction distribution (simulated)
            st.subheader("Prediction Distribution (Simulated):")
            prediction_counts = np.random.randint(1, 100, size=4)
            prediction_counts[class_labels.index(predicted_label)] += 100  # Simulate the current prediction
            
            fig, ax = plt.subplots()
            ax.bar(class_labels, prediction_counts, color=['red', 'green', 'blue', 'orange'])
            ax.set_xlabel('Class')
            ax.set_ylabel('Count')
            ax.set_title('Prediction Distribution')
            st.pyplot(fig)

if __name__ == '__main__':
    main()