File size: 6,269 Bytes
adce462 eab12d2 adce462 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 |
from flask import Flask, render_template, request, jsonify
import torch
import torch.nn as nn
from PIL import Image
import numpy as np
import io
import os
import pickle
from scipy.stats import entropy
app = Flask(__name__)
app.config['MAX_CONTENT_LENGTH'] = 5 * 1024 * 1024
# --------------------
# Model definitions
# --------------------
class BinaryDNN(nn.Module):
def __init__(self):
super().__init__()
self.net = nn.Sequential(
nn.Linear(196614, 384), nn.BatchNorm1d(384), nn.ReLU(), nn.Dropout(0.45),
nn.Linear(384, 192), nn.BatchNorm1d(192), nn.ReLU(), nn.Dropout(0.35),
nn.Linear(192, 96), nn.BatchNorm1d(96), nn.ReLU(), nn.Dropout(0.25),
nn.Linear(96, 2))
def forward(self, x): return self.net(x)
class MultiDNN(nn.Module):
def __init__(self):
super().__init__()
self.net = nn.Sequential(
nn.Linear(196614, 384), nn.BatchNorm1d(384), nn.ReLU(), nn.Dropout(0.45),
nn.Linear(384, 192), nn.BatchNorm1d(192), nn.ReLU(), nn.Dropout(0.35),
nn.Linear(192, 96), nn.BatchNorm1d(96), nn.ReLU(), nn.Dropout(0.25),
nn.Linear(96, 4))
def forward(self, x): return self.net(x)
# --------------------
# Load artifacts ONCE
# --------------------
with open('final_scaler.pkl', 'rb') as f:
scaler = pickle.load(f)
with open('final_bin_le.pkl', 'rb') as f:
binary_le = pickle.load(f)
with open('final_multi_le.pkl', 'rb') as f:
multi_le = pickle.load(f)
device = torch.device('cpu')
binary_model = BinaryDNN()
binary_model.load_state_dict(torch.load('final_bin_best.pth', map_location=device))
binary_model.eval()
multi_model = MultiDNN()
multi_model.load_state_dict(torch.load('final_multi_best.pth', map_location=device))
multi_model.eval()
# --------------------
# V3 Feature extraction
# --------------------
def extract_lsb_feature(img, k=2):
mask = (1 << k) - 1
return (img & mask).flatten()
def extract_randomness_features(lsb_feat):
chunk_size = 512
lsb_reshaped = lsb_feat.reshape(-1, chunk_size)
ent_values = []
for block in lsb_reshaped[::10]:
hist = np.bincount(block.astype(np.int32), minlength=4)
hist = hist / max(hist.sum(), 1)
ent = entropy(hist + 1e-10)
ent_values.append(ent)
mean_entropy = np.mean(ent_values)
sample = lsb_feat[:5000].astype(np.int32)
transitions = np.sum(np.abs(np.diff(sample)))
transition_rate = transitions / max(len(sample) - 1, 1)
hist_full = np.bincount(lsb_feat.astype(np.int32), minlength=4)
expected = len(lsb_feat) / 4
chi2_stat = np.sum((hist_full - expected) ** 2 / expected) if expected > 0 else 0
return np.array([mean_entropy, transition_rate, chi2_stat], dtype=np.float32)
def extract_url_pattern_features(lsb_feat):
chunk_size = 4
num_chunks = len(lsb_feat) // chunk_size
lsb_trimmed = lsb_feat[:num_chunks * chunk_size].astype(np.int32)
byte_values = []
for i in range(0, len(lsb_trimmed), chunk_size):
val = (lsb_trimmed[i] << 6) | (lsb_trimmed[i+1] << 4) | (lsb_trimmed[i+2] << 2) | lsb_trimmed[i+3]
byte_values.append(val)
byte_values = np.array(byte_values)
url_char_count = np.sum((byte_values >= 10) & (byte_values <= 15))
html_char_count = np.sum((byte_values >= 0) & (byte_values <= 5))
unique_ratio = len(np.unique(byte_values)) / max(len(byte_values), 1)
return np.array([
url_char_count / max(len(byte_values), 1),
html_char_count / max(len(byte_values), 1),
unique_ratio
], dtype=np.float32)
def extract_features(image, k=2):
image = image.resize((256, 256), Image.LANCZOS)
if image.mode != 'RGB':
image = image.convert('RGB')
img_array = np.array(image)
lsb_feat = extract_lsb_feature(img_array, k=k)
rand_feat = extract_randomness_features(lsb_feat)
url_feat = extract_url_pattern_features(lsb_feat)
combined = np.concatenate([lsb_feat.astype(np.float32), rand_feat, url_feat])
return combined
# --------------------
# Flask endpoints
# --------------------
@app.route('/')
def index():
return render_template('index.html')
@app.route('/predict', methods=['POST'])
def predict():
try:
file = request.files['file']
image = Image.open(io.BytesIO(file.read()))
features = extract_features(image)
features_normalized = scaler.transform(features.reshape(1, -1)).astype(np.float32)
features_tensor = torch.FloatTensor(features_normalized)
with torch.no_grad():
binary_output = binary_model(features_tensor)
binary_probs = torch.softmax(binary_output, dim=1).cpu().numpy()[0]
binary_pred_idx = int(np.argmax(binary_probs))
binary_class = binary_le.classes_[binary_pred_idx]
if binary_class == 'stego':
with torch.no_grad():
multi_output = multi_model(features_tensor)
multi_probs = torch.softmax(multi_output, dim=1).cpu().numpy()[0]
multi_pred_idx = int(np.argmax(multi_probs))
payload_type = multi_le.classes_[multi_pred_idx]
else:
payload_type = 'clean'
multi_probs = np.zeros(len(multi_le.classes_))
multi_probs[multi_le.transform(['clean'])[0]] = 1.0
result = {
'binary_prediction': {
'class': binary_class,
'confidence': float(binary_probs[binary_pred_idx]),
'probabilities': {c: float(binary_probs[binary_le.transform([c])[0]]) for c in binary_le.classes_}
},
'multiclass_prediction': {
'payload_type': payload_type,
'probabilities': {c: float(multi_probs[i]) for i, c in enumerate(multi_le.classes_)}
}
}
return jsonify(result)
except Exception as e:
import traceback
print('Error:', str(e))
traceback.print_exc()
return jsonify({'error': str(e)}), 500
@app.errorhandler(413)
def request_entity_too_large(error):
return jsonify({'error': 'File size exceeds 5MB'}), 413
if __name__ == "__main__":
port = int(os.environ.get("PORT", 7860))
app.run(debug=True, host="0.0.0.0", port=port) |