File size: 15,479 Bytes
27f6252 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 | #!/usr/bin/env python3
"""
Hugging Face Inference Engine for CVE Cybersecurity LLM
Uses Llama 3.1 8B model for CVE analysis and security recommendations.
Supports both base model and fine-tuned models.
"""
import json
import logging
import time
import os
from pathlib import Path
from typing import Dict, List, Any, Optional
import sys
# Add project root to path
project_root = Path(__file__).parent.parent.parent
sys.path.insert(0, str(project_root))
from config import Config
# Import transformers
try:
from transformers import AutoTokenizer, AutoModelForCausalLM
from peft import PeftModel, PeftConfig
import torch
from torch.nn import functional as F
from huggingface_hub import login
except ImportError as e:
print(f"Error importing transformers: {e}")
print("Please install: pip install transformers torch accelerate peft")
sys.exit(1)
# Configure logging
logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(levelname)s - %(message)s')
logger = logging.getLogger(__name__)
class HFCVEInferenceEngine:
"""Hugging Face inference engine for CVE cybersecurity analysis"""
def __init__(self, model_name: str = "meta-llama/Meta-Llama-3-8B", use_fine_tuned: bool = False, fine_tuned_path: str = "models/fine_tuned_cve_production"):
self.config = Config()
self.model_name = model_name
self.device = "cuda" if torch.cuda.is_available() else "cpu"
self.use_fine_tuned = use_fine_tuned
self.fine_tuned_path = Path(fine_tuned_path)
# Load and authenticate with token
self.token = self._load_token()
self._authenticate()
# Initialize model and tokenizer
self.tokenizer = None
self.model = None
# System prompts for different analysis types
self.system_prompts = {
'vulnerability_analysis': """You are a cybersecurity expert. Analyze the CVE and provide:
- Severity assessment (Critical/High/Medium/Low)
- Key technical details
- Impact analysis
- Brief remediation steps
Keep responses concise and actionable.""",
'remediation': """You are a cybersecurity specialist. Provide specific remediation steps:
- Immediate actions
- Patch information
- Workarounds
- Verification steps
Focus on practical solutions.""",
'technical_analysis': """You are a technical security analyst. Provide:
- Root cause analysis
- Attack mechanism
- Technical impact
- Detection recommendations
Use technical depth for security professionals.""",
'product_analysis': """You are a product security analyst. Analyze:
- Vulnerability trends
- Risk assessment
- Security recommendations
- Vendor practices
Provide actionable insights."""
}
logger.info(f"Initializing HF inference engine on {self.device}")
if self.use_fine_tuned:
logger.info(f"Using fine-tuned model from: {self.fine_tuned_path}")
else:
logger.info(f"Using base model: {self.model_name}")
self._load_model()
def _load_token(self) -> str:
"""Load Hugging Face token from file"""
token_file = Path("llama_token.txt")
if token_file.exists():
with open(token_file, 'r') as f:
token = f.read().strip()
logger.info("✅ Hugging Face token loaded")
return token
else:
raise FileNotFoundError("llama_token.txt not found. Please create this file with your Hugging Face token.")
def _authenticate(self):
"""Authenticate with Hugging Face"""
try:
# Set environment variable for token
os.environ["HF_TOKEN"] = self.token
login(token=self.token)
logger.info("✅ Successfully authenticated with Hugging Face")
except Exception as e:
logger.error(f"❌ Authentication failed: {e}")
raise
def _load_model(self):
"""Load the model and tokenizer"""
try:
if self.use_fine_tuned:
self._load_fine_tuned_model()
else:
self._load_base_model()
except Exception as e:
logger.error(f"Error loading model: {e}")
raise
def _load_fine_tuned_model(self):
"""Load fine-tuned model with LoRA adapter"""
if not self.fine_tuned_path.exists():
logger.error(f"Fine-tuned model not found at {self.fine_tuned_path}")
logger.info("Falling back to base model...")
self.use_fine_tuned = False
self._load_base_model()
return
logger.info(f"Loading fine-tuned model from {self.fine_tuned_path}")
try:
# Load LoRA configuration
config = PeftConfig.from_pretrained(self.fine_tuned_path)
logger.info(f"Base model: {config.base_model_name_or_path}")
logger.info(f"Adapter type: {config.peft_type}")
# Load tokenizer from fine-tuned model
logger.info("Loading tokenizer...")
self.tokenizer = AutoTokenizer.from_pretrained(self.fine_tuned_path)
if self.tokenizer.pad_token is None:
self.tokenizer.pad_token = self.tokenizer.eos_token
# Load base model
logger.info("Loading base model...")
base_model = AutoModelForCausalLM.from_pretrained(
config.base_model_name_or_path,
token=self.token,
torch_dtype=torch.float16 if self.device == "cuda" else torch.float32,
device_map="auto" if self.device == "cuda" else None,
trust_remote_code=True
)
# Load LoRA adapter
logger.info("Loading LoRA adapter...")
self.model = PeftModel.from_pretrained(base_model, self.fine_tuned_path)
if self.device == "cpu":
self.model = self.model.to(self.device)
logger.info("✅ Fine-tuned model loaded successfully!")
except Exception as e:
logger.error(f"Failed to load fine-tuned model: {e}")
logger.info("Falling back to base model...")
self.use_fine_tuned = False
self._load_base_model()
def _load_base_model(self):
"""Load base model"""
logger.info(f"Loading tokenizer for {self.model_name}...")
self.tokenizer = AutoTokenizer.from_pretrained(
self.model_name,
token=self.token,
trust_remote_code=True
)
logger.info(f"Loading model for {self.model_name}...")
self.model = AutoModelForCausalLM.from_pretrained(
self.model_name,
token=self.token,
torch_dtype=torch.float16 if self.device == "cuda" else torch.float32,
device_map="auto" if self.device == "cuda" else None,
trust_remote_code=True
)
if self.device == "cpu":
self.model = self.model.to(self.device)
logger.info("✅ Base model and tokenizer loaded successfully")
def generate_response(self, query: str, context: List[Dict[str, Any]],
analysis_type: str = 'vulnerability_analysis') -> str:
"""Generate response using the loaded model"""
if not self.model or not self.tokenizer:
return "Error: Model not loaded"
try:
# Build context text
context_text = self._build_context_text(context)
# Choose prompt format based on model type
if self.use_fine_tuned:
# Use instruction format for fine-tuned model
full_prompt = self._format_prompt_for_fine_tuned(query, context_text, analysis_type)
else:
# Use system prompt format for base model
full_prompt = self._format_prompt_for_base_model(query, context_text, analysis_type)
# Tokenize
inputs = self.tokenizer(full_prompt, return_tensors="pt", truncation=True, max_length=2048)
inputs = {k: v.to(self.device) for k, v in inputs.items()}
# Generate
with torch.no_grad():
outputs = self.model.generate(
**inputs,
max_new_tokens=512,
temperature=0.3,
top_p=0.9,
do_sample=True,
pad_token_id=self.tokenizer.eos_token_id
)
# Decode response and clean up
response = self.tokenizer.decode(outputs[0][inputs['input_ids'].shape[1]:], skip_special_tokens=True)
# Clean up any artifacts from the model output
response = response.replace('<|eot_id|>', '').replace('ocausticassist', '').strip()
return response
except Exception as e:
logger.error(f"Error generating response: {e}")
return f"Error: {str(e)}"
def _format_prompt_for_fine_tuned(self, query: str, context_text: str, analysis_type: str) -> str:
"""Format prompt for fine-tuned model using instruction format"""
# Map analysis type to instruction
instruction_map = {
'vulnerability_analysis': "Analyze this CVE vulnerability and provide a comprehensive security assessment.",
'remediation': "Provide comprehensive remediation guidance with MITRE ATT&CK mitigations for this CVE vulnerability.",
'technical_analysis': "Provide a detailed technical analysis with MITRE ATT&CK context for this CVE vulnerability.",
'detection': "Create comprehensive detection strategies for this CVE vulnerability.",
'mitigation': "Develop a comprehensive mitigation strategy for this CVE vulnerability."
}
instruction = instruction_map.get(analysis_type, "Analyze this cybersecurity query and provide a comprehensive response.")
return f"""### Instruction:
{instruction}
### Input:
CVE Information:
{context_text}
Question: {query}
### Response:
"""
def _format_prompt_for_base_model(self, query: str, context_text: str, analysis_type: str) -> str:
"""Format prompt for base model using system prompt format"""
# Get system prompt
system_prompt = self.system_prompts.get(analysis_type, self.system_prompts['vulnerability_analysis'])
# Build full prompt for Llama 3
return f"""<|begin_of_text|><|start_header_id|>system<|end_header_id|>
{system_prompt}<|eot_id|><|start_header_id|>user<|end_header_id|>
CVE Information:
{context_text}
Question: {query}<|eot_id|><|start_header_id|>assistant<|end_header_id|>
"""
def _build_context_text(self, context: List[Dict[str, Any]]) -> str:
"""Build context text from CVE data"""
if not context:
return "No CVE data provided."
context_parts = []
for item in context:
if isinstance(item, dict):
if 'cve_id' in item:
context_parts.append(f"CVE ID: {item['cve_id']}")
if 'description' in item:
context_parts.append(f"Description: {item['description']}")
if 'cvss_v3' in item and 'base_score' in item['cvss_v3']:
context_parts.append(f"CVSS Score: {item['cvss_v3']['base_score']}")
if 'products' in item:
products = ", ".join(item['products'][:5]) # Limit to 5 products
context_parts.append(f"Affected Products: {products}")
return "\n".join(context_parts)
def analyze_cve(self, cve_id: str, cve_data: Dict[str, Any]) -> str:
"""Analyze a specific CVE"""
context = [cve_data]
return self.generate_response(f"Analyze CVE {cve_id}", context, 'vulnerability_analysis')
def get_remediation_plan(self, cve_id: str, cve_data: Dict[str, Any]) -> str:
"""Get remediation plan for a CVE"""
context = [cve_data]
return self.generate_response(f"Provide remediation plan for CVE {cve_id}", context, 'remediation')
def analyze_product_security(self, product_name: str, cve_list: List[Dict[str, Any]]) -> str:
"""Analyze security for a specific product"""
context = cve_list[:5] # Limit to top 5 CVEs
return self.generate_response(f"Analyze security for {product_name}", context, 'product_analysis')
def main():
"""Main function to test the inference engine"""
import argparse
parser = argparse.ArgumentParser(description="CVE LLM Inference Engine")
parser.add_argument("--use_fine_tuned", action="store_true",
help="Use fine-tuned model instead of base model")
parser.add_argument("--fine_tuned_path", default="models/fine_tuned_cve_production",
help="Path to fine-tuned model")
parser.add_argument("--query", default="Analyze CVE-2021-44228 Log4j vulnerability",
help="Test query to run")
args = parser.parse_args()
try:
# Initialize inference engine
engine = HFCVEInferenceEngine(
use_fine_tuned=args.use_fine_tuned,
fine_tuned_path=args.fine_tuned_path
)
# Test with sample CVE data
sample_cve = {
'cve_id': 'CVE-2021-44228',
'description': 'Apache Log4j2 2.0-beta9 through 2.14.1 JNDI features used in configuration, log messages, and parameters do not protect against attacker controlled LDAP and other JNDI related endpoints.',
'cvss_v3': {'base_score': 10.0, 'base_severity': 'CRITICAL'},
'products': ['Apache Log4j2']
}
# Test different analysis types
print("\n" + "="*60)
print("TESTING CVE ANALYSIS")
print("="*60)
# Vulnerability analysis
print("\n1. VULNERABILITY ANALYSIS:")
print("-" * 40)
response = engine.analyze_cve('CVE-2021-44228', sample_cve)
print(response)
# Remediation plan
print("\n2. REMEDIATION PLAN:")
print("-" * 40)
response = engine.get_remediation_plan('CVE-2021-44228', sample_cve)
print(response)
# Custom query
print("\n3. CUSTOM QUERY:")
print("-" * 40)
response = engine.generate_response(args.query, [sample_cve], 'vulnerability_analysis')
print(response)
print(f"\n✅ Inference engine test completed successfully!")
print(f"Model type: {'Fine-tuned' if args.use_fine_tuned else 'Base'}")
except Exception as e:
logger.error(f"❌ Inference engine test failed: {e}")
raise
if __name__ == "__main__":
main() |