Multiple DNS Cache poisoning vulnerabilities in dproxy and drpoxy-nexgen (CVE-2022-33988, CVE-2022-33989, CVE-2022-33990, CVE-2022-33991)

["Philipp Jeitner (SIT)" <philipp.jeitner () sit fraunhofer de>]

We hereby disclose the discovery of multiple DNS Cache poisoning 
vulnerabilities in the dproxy(-nexgen) DNS forwarder. dproxy is a 
caching DNS forwarder/proxy which is unmaintained since about 2004, yet 
it is still used in some residential router firmwares. Because the 
project is unmaintained, there are no patches available for the 
described issues.

Our findings are published in our 2022 paper "XDRI Attacks - and - How 
to Enhance Resilience of Residential Routers" in August 2022.

Discovery/Credits
-----------------

Philipp Jeitner, Lucas Teichmann and Haya Shulman
Fraunhofer SIT

References
----------

  - dproxy: https://sourceforge.net/projects/dproxy/
  - paper website: https://xdi-attack.net/
  - paper presentation: 
https://www.usenix.org/conference/usenixsecurity22/presentation/jeitner



CVE-2022-33990: Misinterpretation of special characters in domain names 
leading to cache-poisoning
--------------------------------------------------------------------------------------------------

Misinterpretation of special domain name characters in dproxy-nexgen 
leads to cache-poisoning as domain names and their associated IP 
addresses are cached in their misinterpreted form.

## Summary

Attacker can poison the DNS cache of the vulnerable router/forwarder by 
triggering queries to attacker controlled domain names whose queries 
and/or answers contain special characters (zero-byte or period sign). 
These characters are misinterpreted by the vulnerable router/forwarder 
so that the attacker can provide addresses for domain names he does not own.

## Impact

Attackers who control a script or web-site which is loaded on a client 
of the vulnerable router/forwarder can hijack connections by poisoning 
the DNS cache.

## Steps to reproduce

To reproduce, connect a computer to the router and follow the Steps at 
https://xdi-attack.net/manual.html or use our downloadable test-tool at 
https://xdi-attack.net/test.html (NOT the online test).

## Detailed description and publication timeline

A detailed description of this attack is included in our 2021 USENIX 
security paper "Injection Attacks Reloaded: Tunnelling Malicious 
Payloads over DNS", see Section 3.2. We conducted further research and 
found that these attacks apply to various router models.



CVE-2022-33989: Static UDP port in DNS queries sent to upstream resolvers
-------------------------------------------------------------------------

dproxy-nexgen uses a static UDP source port (selected randomly only at 
boot-time) in upstream queries sent to DNS resolvers which allows DNS 
cache poisoning as there is not enough entropy to prevent traffic 
injection attacks.

## Summary

The router/forwarder uses a fixed UDP port for all queries sent to 
upstream resolvers.

## Impact

Attackers who control a script or web-site which is loaded on a client 
of the vulnerable router/forwarder can exploit this to poison the DNS 
cache by classic DNS poisoning attacks with spoofed IP address of the 
upstream resolver.

## Steps to reproduce

Connect a computer to the vulnerable router/forwarder and trigger 
multiple DNS queries. Observe the queries sent to upstream resolvers via 
packet capture, either on the routers Internet-facing interface or the 
upstream resolver's network interface. The queries captured on these 
interfaces have the same UDP source port.

## Detailed description and publication timeline

This attack is known to be practical since the 2008 publication "Black 
Ops 2008: It’s The End Of The Cache As We Know It" 
(https://www.blackhat.com/presentations/bh-jp-08/bh-jp-08-Kaminsky/BlackHat-Japan-08-Kaminsky-DNS08-BlackOps.pdf). 
During an evaluation of DNS vulnerabilities in routers, we found this 
attack to be still applicable.



CVE-2022-33988: DNS TXID value is re-used from client queries
-------------------------------------------------------------

dproxy-nexgen re-uses the DNS transaction id (TXID) value from client 
queries, which allows attackers able to sent queries to the resolver to 
conduct DNS cache poisoning attacks as the TXID value is known to the 
attacker.

## Summary

The router/forwarder re-uses the DNS TXID value from client queries when 
forwarding the DNS query to upstream resolvers.

## Impact

Attackers which can send DNS queries directly to the vulnerable 
router/forwarder can infer the TXID used for upstream queries as the 
value is re-used from the query sent by the attacker. This allows the 
attacker to inject malicious records into the router/forwarder's DNS 
cache, as the remaining entropy in the DNS request (max 16-bit UDP 
source port) is not enough to protect against spoofed DNS responses.

## Steps to reproduce

Connect a computer to the vulnerable router/forwarder and trigger some 
DNS queries. Monitor the requests sent by the client device (e.g. via 
wireshark) and compare the DNS TXID value with the value of the packets 
forwarded to the upstream resolvers, e.g. by capturing traffic on the 
routers Internet-facing interface or the upstream resolver's network 
interface. The TXID values of these packets are the same.



CVE-2022-33991: Disabling of DNSSEC protection provided by upstream 
resolvers
-----------------------------------------------------------------------------

dproxy-nexgen forwards and caches DNS queries with checking disabled 
(CD) bit set to 1 which leads to disabling of DNSSEC protection provided 
by upstream resolvers.

## Summary

The router/forwarder forwards DNS queries with the checking disabled 
(CD) bit set to 1 to upstream resolvers and caches the responses 
provided by the upstream resolver. The cached answers are then sent to 
other clients even when they do set the checking disabled (CD) bit to 0.

## Impact

Attackers which can send DNS queries directly to the vulnerable 
router/forwarder can disable DNSSEC protection on the upstream resolver 
by sending queries with the checking disabled (CD) bit set to 1. When 
the attacker is able to inject DNS responses via another method (e.g. 
MitM attacks, BGP hijacking), this allows attacker to hijack connections 
from clients of the vulnerable router/forwarder, as DNSSEC protection is 
not guaranteed anymore.

## Steps to reproduce

Connect a computer to the vulnerable router/forwarder and trigger the 
following DNS queries via `dig`:

     $ dig sigfail.verteiltesysteme.net +cdflag @router/forwarder-ip
     (should always return 134.91.78.139)

     $ dig sigfail.verteiltesysteme.net +short @router/forwarder-ip
     (returns 134.91.78.139 if vulnerable, should return nothing)

Note: you can replace `sigfail.verteiltesysteme.net` with any other 
domain with broken DNSSEC, such as `www.dnssec-failed.org`, only the 
addresses will be different.