File size: 14,380 Bytes
5a2c9e0 59dc700 5a2c9e0 59dc700 5a2c9e0 59dc700 5a2c9e0 59dc700 5a2c9e0 59dc700 5a2c9e0 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 | """
Rhodawk AI — Autonomous Fuzzing Engine
=======================================
Generates language-aware fuzzing harnesses using LLM then executes them.
Integrates with AFL++, libFuzzer (via atheris for Python), and Hypothesis.
Pipeline per target:
1. LLM generates a harness tailored to the target function/API
2. Harness is written to /tmp and compiled/instrumented
3. Fuzzer runs for duration_s seconds with coverage feedback
4. Crashes are triaged: unique crashes extracted, deduped by stack hash
5. Results returned for exploit_primitives reasoning
Supported modes:
- Python → atheris (libFuzzer bindings for Python)
- C/C++ → AFL++ subprocess (if installed)
- JS/TS → jsfuzz / fast-check property testing
- Generic → Hypothesis with AI-generated strategies
"""
from __future__ import annotations
import hashlib
import json
import os
import subprocess
import tempfile
import time
from dataclasses import dataclass, field
from typing import Optional
import requests
OPENROUTER_API_KEY = os.getenv("OPENROUTER_API_KEY", "")
FUZZ_MODEL = os.getenv("HERMES_FAST_MODEL", "deepseek/deepseek-v3:free")
OPENROUTER_BASE = "https://openrouter.ai/api/v1"
MAX_FUZZ_DURATION = int(os.getenv("RHODAWK_MAX_FUZZ_DURATION", "120"))
FUZZ_CORPUS_DIR = os.getenv("RHODAWK_FUZZ_CORPUS", "/data/fuzz_corpus")
@dataclass
class CrashRecord:
crash_id: str
target: str
crash_input: str
crash_output: str
stack_hash: str
crash_type: str # segfault | assertion | exception | timeout | oom
is_unique: bool
reproducer_path: str
timestamp: str = field(default_factory=lambda: time.strftime("%Y-%m-%dT%H:%M:%SZ"))
@dataclass
class FuzzResult:
target: str
language: str
duration_s: int
total_executions: int
unique_crashes: list[CrashRecord]
coverage_percent: float
harness_code: str
error: Optional[str] = None
def _llm_generate_harness(
target: str,
language: str,
repo_dir: str,
source_context: str,
) -> str:
"""Use LLM to generate a fuzzing harness for the target function."""
if not OPENROUTER_API_KEY:
return _fallback_harness(target, language)
system = (
"You are an expert fuzzing engineer. Generate a minimal, correct fuzzing harness "
"for the given target. The harness must: "
"1) Accept raw bytes as input, 2) Parse them into valid arguments, "
"3) Call the target without crashing on invalid input (catch exceptions), "
"4) Be as fast as possible (no I/O, no sleep). "
"Return ONLY the harness code, no explanation."
)
if language == "python":
prompt = (
f"TARGET FUNCTION: {target}\n"
f"LANGUAGE: Python (atheris/libFuzzer)\n"
f"SOURCE CONTEXT:\n```python\n{source_context[:2000]}\n```\n\n"
"Generate an atheris fuzzing harness. Import atheris and the target module. "
"The TestOneInput function must accept bytes. Use FuzzedDataProvider to extract typed values."
)
elif language in ("javascript", "typescript"):
prompt = (
f"TARGET FUNCTION: {target}\n"
f"LANGUAGE: {language} (jsfuzz)\n"
f"SOURCE CONTEXT:\n```javascript\n{source_context[:2000]}\n```\n\n"
"Generate a jsfuzz harness. Export a default async function that accepts Buffer."
)
elif language in ("go",):
prompt = (
f"TARGET FUNCTION: {target}\n"
f"LANGUAGE: Go (native fuzzing)\n"
f"SOURCE CONTEXT:\n```go\n{source_context[:2000]}\n```\n\n"
"Generate a Go fuzz test using testing.F and f.Fuzz()."
)
else:
prompt = (
f"TARGET: {target}\n"
f"LANGUAGE: {language}\n"
f"SOURCE:\n```\n{source_context[:2000]}\n```\n\n"
"Generate a Hypothesis property-based test that explores the target's input space."
)
headers = {
"Authorization": f"Bearer {OPENROUTER_API_KEY}",
"Content-Type": "application/json",
}
payload = {
"model": FUZZ_MODEL,
"messages": [
{"role": "system", "content": system},
{"role": "user", "content": prompt},
],
"temperature": 0.1,
"max_tokens": 1500,
}
try:
resp = requests.post(
f"{OPENROUTER_BASE}/chat/completions",
headers=headers, json=payload, timeout=60,
)
resp.raise_for_status()
content = resp.json()["choices"][0]["message"]["content"]
code = _extract_code_block(content)
return code or _fallback_harness(target, language)
except Exception:
return _fallback_harness(target, language)
def _extract_code_block(text: str) -> str:
"""Extract first code block from markdown."""
import re
m = re.search(r"```(?:python|javascript|go|typescript|rust)?\n(.*?)```", text, re.DOTALL)
if m:
return m.group(1).strip()
lines = [l for l in text.splitlines() if not l.startswith("```")]
return "\n".join(lines).strip()
def _fallback_harness(target: str, language: str) -> str:
"""Generic fallback harness when LLM is unavailable."""
if language == "python":
# FIX (Build Error): atheris requires Clang + libFuzzer at compile time and
# fails to build on HuggingFace Spaces. The fallback harness now uses
# hypothesis which is available everywhere, matching the Hypothesis
# fallback already in use when atheris is unavailable at runtime.
return f"""
import sys
try:
import atheris
_ATHERIS_AVAILABLE = True
except ImportError:
_ATHERIS_AVAILABLE = False
if _ATHERIS_AVAILABLE:
import sys
@atheris.instrument_func
def fuzz_target(data):
fdp = atheris.FuzzedDataProvider(data)
try:
val = fdp.ConsumeUnicodeNoSurrogates(128)
# TODO: call {target}(val)
except Exception:
pass
atheris.Setup(sys.argv, fuzz_target)
atheris.Fuzz()
else:
# Hypothesis-based fallback when atheris/libFuzzer is unavailable
from hypothesis import given, settings, HealthCheck
from hypothesis import strategies as st
@given(st.text(max_size=128))
@settings(max_examples=500, suppress_health_check=list(HealthCheck))
def fuzz_target(val):
try:
# TODO: call {target}(val)
pass
except Exception:
pass
fuzz_target()
"""
return f"# Fallback harness for {target} ({language})\n# Manual harness required\n"
def _get_source_context(repo_dir: str, target: str, language: str) -> str:
"""Extract relevant source code context around the target function."""
import glob as _glob
ext_map = {
"python": ["*.py"], "javascript": ["*.js"], "typescript": ["*.ts"],
"go": ["*.go"], "rust": ["*.rs"], "java": ["*.java"], "ruby": ["*.rb"],
}
extensions = ext_map.get(language, ["*.*"])
for ext in extensions:
for fpath in _glob.glob(f"{repo_dir}/**/{ext}", recursive=True):
if "test" in fpath.lower() or "node_modules" in fpath:
continue
try:
content = open(fpath).read()
if target.split(".")[-1] in content or target.split("::")[-1] in content:
rel = os.path.relpath(fpath, repo_dir)
return f"# File: {rel}\n{content[:3000]}"
except Exception:
pass
return f"# Could not find source for {target}"
def _run_python_atheris(harness_code: str, duration_s: int) -> list[CrashRecord]:
"""Run atheris fuzzer on a Python harness."""
crashes = []
harness_path = None
corpus_dir = None
try:
fd, harness_path = tempfile.mkstemp(suffix="_fuzz.py")
with os.fdopen(fd, "w") as f:
f.write(harness_code)
corpus_dir = tempfile.mkdtemp(prefix="fuzz_corpus_")
seed_file = os.path.join(corpus_dir, "seed")
with open(seed_file, "wb") as f:
f.write(b"hello world\x00\xff")
crash_dir = tempfile.mkdtemp(prefix="fuzz_crashes_")
cmd = [
"python", harness_path,
f"-max_total_time={duration_s}",
f"-artifact_prefix={crash_dir}/",
corpus_dir,
]
proc = subprocess.Popen(
cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE,
text=True, timeout=duration_s + 30,
)
try:
stdout, stderr = proc.communicate(timeout=duration_s + 30)
except subprocess.TimeoutExpired:
proc.kill()
stdout, stderr = proc.communicate()
combined = stdout + stderr
for crash_file in os.listdir(crash_dir):
if crash_file.startswith("crash-") or crash_file.startswith("oom-"):
crash_path = os.path.join(crash_dir, crash_file)
try:
with open(crash_path, "rb") as f:
crash_bytes = f.read()
crash_input = crash_bytes.hex()[:500]
stack_hash = hashlib.sha256(crash_bytes[:64]).hexdigest()[:16]
crashes.append(CrashRecord(
crash_id=hashlib.sha256(crash_bytes).hexdigest()[:12],
target="python_harness",
crash_input=crash_input,
crash_output=combined[-1000:],
stack_hash=stack_hash,
crash_type="exception" if "crash-" in crash_file else "oom",
is_unique=True,
reproducer_path=crash_path,
))
except Exception:
pass
except Exception as e:
crashes.append(CrashRecord(
crash_id="setup_error",
target="python_harness",
crash_input="",
crash_output=str(e),
stack_hash="error",
crash_type="setup_error",
is_unique=False,
reproducer_path="",
))
finally:
if harness_path and os.path.exists(harness_path):
os.unlink(harness_path)
return crashes
def _run_hypothesis(repo_dir: str, target: str, harness_code: str, duration_s: int) -> list[CrashRecord]:
"""Run Hypothesis property-based testing as a fuzzing fallback."""
crashes = []
fd, test_path = tempfile.mkstemp(suffix="_hyp_test.py", dir="/tmp")
try:
with os.fdopen(fd, "w") as f:
f.write(harness_code)
proc = subprocess.Popen(
["python", "-m", "pytest", test_path, "-x", "--tb=short",
f"--hypothesis-seed=0", "-q"],
cwd=repo_dir, stdout=subprocess.PIPE, stderr=subprocess.PIPE,
text=True,
)
try:
stdout, stderr = proc.communicate(timeout=min(duration_s, 60))
except subprocess.TimeoutExpired:
proc.kill()
stdout, stderr = proc.communicate()
combined = (stdout or "") + (stderr or "")
if "FAILED" in combined or "AssertionError" in combined or "Falsifying" in combined:
stack_hash = hashlib.sha256(combined[:200].encode()).hexdigest()[:16]
crashes.append(CrashRecord(
crash_id=stack_hash,
target=target,
crash_input="see Hypothesis output",
crash_output=combined[:2000],
stack_hash=stack_hash,
crash_type="assertion",
is_unique=True,
reproducer_path=test_path,
))
except Exception as e:
pass
finally:
try:
os.unlink(test_path)
except OSError:
pass
return crashes
def run_fuzzing_campaign(
repo_dir: str,
target: str,
language: str = "python",
duration_s: int = 60,
) -> dict:
"""
Main entry point. Generate harness + run fuzzer + return triage results.
"""
duration_s = min(duration_s, MAX_FUZZ_DURATION)
print(f"[FUZZ] Starting campaign: {target} ({language}, {duration_s}s)")
source_context = _get_source_context(repo_dir, target, language)
harness_code = _llm_generate_harness(target, language, repo_dir, source_context)
start = time.time()
if language == "python" and "atheris" in harness_code:
# FIX (Build Error): atheris may be unavailable; fall back to Hypothesis.
try:
import importlib.util
_atheris_available = importlib.util.find_spec("atheris") is not None
except Exception:
_atheris_available = False
if _atheris_available:
crashes = _run_python_atheris(harness_code, duration_s)
else:
# Rewrite the harness to use Hypothesis if atheris is not installed
harness_code = harness_code.replace(
"import atheris", "# atheris unavailable — using Hypothesis fallback"
)
crashes = _run_hypothesis(repo_dir, target, harness_code, duration_s)
else:
crashes = _run_hypothesis(repo_dir, target, harness_code, duration_s)
elapsed = time.time() - start
unique_hashes = set()
unique_crashes = []
for c in crashes:
if c.stack_hash not in unique_hashes:
unique_hashes.add(c.stack_hash)
unique_crashes.append(c)
result = FuzzResult(
target=target,
language=language,
duration_s=int(elapsed),
total_executions=len(crashes),
unique_crashes=unique_crashes,
coverage_percent=0.0,
harness_code=harness_code,
)
print(f"[FUZZ] Done: {len(unique_crashes)} unique crash(es) in {elapsed:.1f}s")
return {
"target": result.target,
"language": result.language,
"duration_s": result.duration_s,
"unique_crashes": len(result.unique_crashes),
"harness_code": result.harness_code[:500],
"crashes": [
{
"id": c.crash_id, "type": c.crash_type,
"input_hex": c.crash_input[:100],
"output_snippet": c.crash_output[:500],
"stack_hash": c.stack_hash,
}
for c in result.unique_crashes
],
"has_crashes": len(result.unique_crashes) > 0,
}
|