app.py

import gradio as gr
import pandas as pd
import random
from sklearn.ensemble import IsolationForest

# Global login history
login_history = []

# Simulate typing speed (chars/sec)
def simulate_typing_speed(normal=True):
    return round(random.uniform(3.0, 6.0), 2) if normal else round(random.uniform(0.5, 2.5), 2)

# Simulate login attempt
def simulate_login(normal=True):
    login_time = random.randint(8, 18) if normal else random.choice([0, 23])
    actions = random.randint(5, 20) if normal else random.randint(25, 100)
    typing_speed = simulate_typing_speed(normal)
    # For demo: simulate device hash and IP
    device = "device_1" if normal else random.choice(["device_2", "device_3"])
    ip = "192.168.0.1" if normal else random.choice(["10.0.0.1", "172.16.0.1"])
    return {"time": login_time, "actions": actions, "typing_speed": typing_speed, "device": device, "ip": ip}

# Risk scoring function
def evaluate_risk(new_login, history):
    # Default matches
    device_match = any(l["device"] == new_login["device"] for l in history)
    ip_match = any(l["ip"] == new_login["ip"] for l in history)

    # Anomaly detection only on numeric features
    if len(history) >= 10:
        df = pd.DataFrame(history + [new_login])
        model = IsolationForest(contamination=0.1)
        model.fit(df[["time", "actions", "typing_speed"]])
        pred = model.predict(df[["time", "actions", "typing_speed"]])
        is_anomaly = pred[-1] == -1
    else:
        is_anomaly = False  # not enough data

    # Risk logic
    if not is_anomaly and device_match and ip_match:
        risk = "Low Risk ✅"
        color = "green"
    elif is_anomaly and device_match and ip_match:
        risk = "Medium Risk ⚠️"
        color = "yellow"
    else:
        risk = "High Risk 🚨"
        color = "red"

    return risk, color

# Main function for Gradio
def login(action):
    global login_history
    new_login = simulate_login(normal=(action=="Normal Login"))
    login_history.append(new_login)

    risk, color = evaluate_risk(new_login, login_history)

    # Convert history to DataFrame for display
    df = pd.DataFrame(login_history)
    df = df.rename(columns={"time":"Login Time","actions":"Actions","typing_speed":"Typing Speed","device":"Device","ip":"IP Address"})

    return df, f"{risk}"

# Gradio interface
iface = gr.Interface(
    fn=login,
    inputs=gr.Radio(["Normal Login","Suspicious Login"], label="Select Login Type"),
    outputs=[gr.Dataframe(headers=["Login Time","Actions","Typing Speed","Device","IP Address"]), 
             gr.Textbox(label="Risk Status")],
    live=True,
    title="🛡️ IdentityGuard AI",
    description="AI-powered behavioral login monitoring with anomaly detection and risk scoring"
)

iface.launch()