Create app.py

app.py

@@ -0,0 +1,57 @@
+
+
+import streamlit as st
+import numpy as np
+from tensorflow.keras.models import load_model
+from tensorflow.keras.preprocessing.sequence import pad_sequences
+from tensorflow.keras.preprocessing.text import Tokenizer
+
+# Constants
+MAX_LENGTH = 100
+TOKENIZER_PATH = "tokenizer.json"  # Assuming the tokenizer is saved separately.
+
+# Load pre-trained model
+@st.cache_resource
+def load_trained_model():
+    return load_model("deep_learning_model.h5")
+
+# Load tokenizer
+@st.cache_resource
+def load_tokenizer():
+    import json
+    from tensorflow.keras.preprocessing.text import tokenizer_from_json
+    with open(TOKENIZER_PATH, "r") as f:
+        tokenizer_data = json.load(f)
+    return tokenizer_from_json(tokenizer_data)
+
+# Preprocessing function
+def preprocess_prompt(prompt, tokenizer, max_length):
+    sequence = tokenizer.texts_to_sequences([prompt])
+    padded_sequence = pad_sequences(sequence, maxlen=max_length)
+    return padded_sequence
+
+# Predict function
+def detect_prompt(prompt, model, tokenizer, max_length):
+    processed_prompt = preprocess_prompt(prompt, tokenizer, max_length)
+    prediction = model.predict(processed_prompt)[0][0]
+    class_label = "Malicious" if prediction >= 0.5 else "Valid"
+    confidence_score = prediction * 100 if prediction >= 0.5 else (1 - prediction) * 100
+    return class_label, confidence_score
+
+# Streamlit App
+st.title("Prompt Injection Detection App")
+st.write("Detect and prevent prompt injection attacks using a deep learning model.")
+
+# Load model and tokenizer
+model = load_trained_model()
+tokenizer = load_tokenizer()
+
+# Input Section
+user_input = st.text_area("Enter a prompt to test:", "")
+if st.button("Detect"):
+    if user_input:
+        label, confidence = detect_prompt(user_input, model, tokenizer, MAX_LENGTH)
+        st.write(f"**Predicted Class:** {label}")
+        st.write(f"**Confidence Score:** {confidence:.2f}%")
+    else:
+        st.warning("Please enter a prompt to test.")