  Steve I'm a PC, so shut up Consultant join:2001-03-10 Yorba Linda, CA
| "Core Internet technology found vulnerable"
This just in: said by MSNBC: Researchers found a serious security flaw that left core Internet technology vulnerable to hackers, prompting a secretive effort by international governments and industry experts in recent weeks to prevent global disruptions of Web surfing, e-mails and instant messages. ... Watson, who runs the www.terrorist.net Web site, predicted that hackers will understand how to begin launching attacks "within five minutes of walking out of that meeting."
-- Stephen J. Friedl * Security Consultant * Tustin, California USA * my web site |
|
 NeenerNeener
join:2002-02-14 | DShield/ISC has extended information including how to turn on MD5 checksums for various BGP implementations »isc.sans.org/diary.php |
|
  Bubba GIT-R-DONE Premium,MVM join:2002-08-19 Around, Us
·Comcast
| reply to Steve Since Spring has Sprung....perhaps now is as good as any to pull the plug.
In a related article from our front page news....»Major TCP Vulnerability Unveiled -- *Team Z* Member |
|
 Daemon Premium join:2003-06-29 San Francisco, CA
·Comcast
| From reading the vulnerability write up, I can image how it could easily be done, and I'm no expert. This looks like a major oversight in the original implementation. Once again, it appears the designers didn't anticipate the maliciousness of the people who would eventually be using the internet. (not that I blame them)
I'm hoping it might spark some sort of internet protocol leap, where we get TCP2, SMTP2, etc, but I know i'm just dreaming. -- -Ryan Find me in the networking and Microsoft help forums |
|
  Sparrow Crystal Sky Premium join:2002-12-03 Sachakhand
| reply to Steve Excellent (and very long) article from April 4, 2004 that addresses many problems we now face: The Rise of Complex Terrorism.
AS Dorothy once said in the Wizard of OZ, "We're not in Kansas anymore, Toto!" -- Security Forum FAQs .. ♥ .. Computer Cops - Symantec Forum .. ♥ .. Starfire "5 in 4" |
|
  Bubba GIT-R-DONE Premium,MVM join:2002-08-19 Around, Us
·Comcast
| reply to Daemon said by Daemon : it appears the designers didn't anticipate the maliciousness of the people who would eventually be using the internet.
Exactly....they were as thoughtful as they could be back in those days....similar to those folks that uttered the words We the people. To bad we couldn't stick by the words spoken from one of our Internet Pioneers and play nice.
"Be liberal in what you accept, and conservative in what you send." -- *Team Z* Member |
|
  Steve I'm a PC, so shut up Consultant join:2001-03-10 Yorba Linda, CA
| reply to Steve new IETF draft dealing with TCP security:
Transmission Control Protocol security considerations -- Stephen J. Friedl * Security Consultant * Tustin, California USA * my web site |
|
  antiserious The Future ain't what it used to be Premium join:2001-12-12 Scranton, PA
| reply to Steve
... it may be time to re-construct The Internet as we know it ... much of it has become corrupt and hostile anyway - not that someone won't try to abuse whatever Next Big Thing comes along, but perhaps if it was done from a security-first POV it might fare better ...
... I'm sure someone, somewhere, is tinkering with that very idea right now ...
... f w i w ... -- ... "It's always been my hope that God has a sense of humor" ... Andy Sipowicz ... |
|
  Morac
join:2001-08-30 Riverside, NJ
·Comcast
| reply to Steve According to this description it looks very simple to to do, but I doubt the home user has much to worry about since all it does is reset the connection. Most TCP connections don't last that long and since it requires knowing the sequence number and receive window size ahead of time it doesn't seem like something a script kiddie could take advantage of. Since it requires a "hands on" touch I doubt anyone will waste their time going after specific home users.
This will mainly affect machines/routers that constantly keep open a large number of TCP connections. Implementing the ACK number checks will make this a lot harder to exploit, which I'm sure is what the higher risk targets have been doing. -- "snmp: the standard e-mail protocol on the Internet" - LinkSys user manual (page 17) |
|
  TivoNut Premium join:2002-04-18 Yorba Linda, CA
| reply to Steve I just took care of one of my BGP upstreams (Verio) and am still working on the other. It was quick and painless.
I sent an email to support at verio.net with my request. They called me about 15 minutes later and scheduled an engineer to coordinate the password addition. Less than 30 minutes after that they called again and we made the change. The circuit bounced and came right back up. Very easy to do.
Verio stated they started receiving requests for this change last week and they're now flooding in. -- "Any information provided to or issued on Internet websites must obtain the inspection and approval of the secrecy censorship." -The Chinese government in a new regulation (January 2000) on future Internet postings. |
|
  major marco Res Firma Mitescere Nescit Premium join:2003-02-13 Stepford, CA clubs:
| reply to Steve The Sky Is Falling Again...
NOT. Don't everyone get yer panties in an uproar over it.
--- From SANS.org
Vulnerability Issues in TCP - Updated; SANS Top-20 Call for Experts The UK National Infrastructure Security Co-Ordination Centre (NISCC) released a vulnerability advisory today on issues in the TCP protocol. The full advisory is available at »www.uniras.gov.uk/vuls/2004/236929/index.htm
According to NISCC, The Border Gateway Protocol (BGP) is judged to be potentially most affected by this vulnerability. BGP relies on a persistent TCP session between BGP peers. Resetting the connection can result in medium term unavailability due to the need to rebuild routing tables and route flapping. Route flapping may result in route dampening (suppression) if the route flaps occur frequently within a short time interval. The overall impact on BGP is likely to be moderate based on the likelihood of successful attack. If the TCP MD5 Signature Option and anti-spoofing measures are used then the impact will be low as these measures will successfully mitigate the vulnerability.
Cisco and Juniper are scheduled to publish announcements about vulnerabilities in the implementation of the BGP code in their respective routers this week. The problem is unauthenticated resets of BGP sessions; a full fix requires replacing the router operating system. The respective advisories will include details on where to get updated operating system images.
A temporary workaround to the problem is to enable MD5 checksums on BGP sessions so that BGP peers can authenticate each other's packets and ignore spoofed TCP resets.
TCP resets can be sent without knowing the exact sequence number used on the TCP connection; by simply landing somewhere in the current TCP window, the RST will succeed.
The SANS Internet Storm Center encourages all BGP enabled Juniper or Cisco router administrators to turn on MD5 checksums as soon as possible while testing the patch supplied by router vendors. A widespread attack on BGP sessions and routing tables has the potential to destabilize the Internet, however most ISP's have already applied the appropriate patches and reduced the probability of such an attack.
Below are instructions on how to enable MD5 checksums for several router platforms, additional details on Unix platforms and routing daemons, and reference links with more information on BGP MD5 checksums.
Cisco IOS configuration:
To enable MD5 checksums between BGP peers, enter the following commands from global configuration mode:
router bgp {your_AS_number} neighbor A.B.C.D password {long_password_value}
So for example, to configure a router that is part of AS 64512 to peer with the router at 10.1.1.2 using the password "a1b2c3d4e5f6g7h8", you would enter:
router bgp 64512 neighbor 10.1.1.2 password a1b2c3d4e5f6g7h8
Please note that the router at 10.1.1.2 would also have to be configured to use the password a1b2c3d4e5f6g7h8 during the peering session as well.
The specific commands to use, including a directive to log when a neighbor joins or leaves a peering arrangement would be:
rtr1#! router 1, local peering address is 10.1.1.1 rtr1#conf term Enter configuration commands, one per line. End with CNTL/Z. rtr1(config)#router bgp 64512 rtr1(config-router)#neighbor 10.1.1.2 password a1b2c3d4e5f6g7h8 rtr1(config-router)#bgp log-neighbor-changes rtr1(config-router)#end rtr1#copy running-config startup-config
The step of copying the running-config to the startup-config needs to be performed at some point. You may wish to wait until the connection is confirmed to be stable before performing this step, which makes the change persist even when the router is powered off.
Juniper router configuration:
On Juniper routers, the
authentication-key a1b2c3d4e5f6g7h8
statement can be used globally, on a particular group, or for an individual BGP peer, using one of the following choices:
edit protocols bgp edit protocols bgp group {groupname} edit protocols bgp group {groupname} neighbor {address}
respectively.
Other platforms:
General-purpose Unix platforms also include BGP peering software in the Zebra and Quagga cross-platform daemons and OpenBSD's bgpd. The MD5 checksum checks need to occur in the operating system kernel, so support for MD5 checksums depends on the operating system kernel as well as on the routing software.
Zebra routing daemon (»www.zebra.org ):
This routing package for Linux, FreeBSD, NetBSD, and OpenBSD does not appear to support RFC2385 MD5 checksums.
Quagga routing daemon (»www.quagga.net ):
According to the documentation, this routing package for Linux, FreeBSD, NetBSD, OpenBSD, and Solaris does not appear to natively support RFC 2385 MD5 checksums. Please see FreeBSD for some patches for that operating system and Quagga.
Bgpd routing daemon:
OpenBSD's bgpd routing daemon supports rfc 2385 checksums.
Tcpdump packet sniffer (»www.tcpdump.org ):
The CVS repository for the tcpdump package includes support for RFC 2385 MD5 checksums via the "-M" command line parameter.
Linux operating system (»www.kernel.org ):
The Linux kernel does not support RFC 2385 MD5 checksums natively. There is support in the form of a kernel patch and additional library at »webchat.webchat.org/~jk/securebgp/
FreeBSD (»www.freebsd.org ):
FreeBSD contains output-only support for RFC 2385. This allows it to connect to a router over a signed connection, but does not accept incoming RFC 2385 connections.
»people.freebsd.org/~bms/dump/quagga-tcpmd5/ »people.freebsd.org/~bms/dump/tcp···rfc2385/
OpenBSD (»www.openbsd.org ):
The OpenBSD kernel in OpenBSD 2.6 and above contains RFC 2385 support.
NetBSD (»www.netbsd.org ):
The NetBSD operating system does not appear to have RFC 2385 support.
Solaris (»www.sun.com ):
The Solaris operating system does not appear to have RFC 2385 support.
General:
Because Cisco implementations only allow 80 character passwords, we recommend you stay under that limit if you either use or peer with any Cisco routers. Juniper routers will go up to 255 characters. The RFC recommends 12 to 24 character passwords. Remember that the password needs to be applied to _both_ peers in a BGP session. If one of the peers does not support RFC 2385, MD5 checksums must be left off for the BGP session between them.
BGP passwords are case sensitive and should not include spaces.
References:
NISCC advisory: »www.uniras.gov.uk/vuls/2004/236929/index.htm
Cisco announcement (none yet, will be updated when available): »www.cisco.com/
Juniper announcement (none yet, will be updated when available): »www.juniper.net/
CERT advisory (none yet, will be updated when available): »www.cert.org/advisories/ »www.us-cert.gov/
CVE entry: »cve.mitre.org/cgi-bin/cvename.cg···004-0230
Configuring BGP on Cisco: »www.cisco.com/univercd/cc/td/doc···p1019545 »www.bind.com/cisco/1cbgp.htm#xtocid2382854
Configuring BGP authentication on Juniper: »www.juniper.net/techpubs/softwar···g20.html »www.qorbit.net/documents/junos-b···late.pdf »www.qorbit.net/documents/junos-b···note.pdf
Global Instability Index, a measure of change in BGP routing tables: »people.ists.dartmouth.edu/~dmcgrath/gii/
Protect BGP sessions with the TCP MD5 option: »bgphints.ruud.org/articles/bgp-md5.html »www.cymru.com/Documents/secure-b···ate.html
RFC 2385, Protection of BGP Sessions via the TCP MD5 Signature Option »www.ietf.org/rfc/rfc2385.txt
Restricting BGP sessions to specific AS numbers on Cisco: »www.cisco.com/en/US/tech/tk365/t···92.shtml
BGP vulnerability discussion:»www.nanog.org/mtg-0306/pdf/franz.pdf
Permanent location for this advisory: »isc.sans.org/diary.php?date=2004-04-20
Credits: - Bill Stearns, the SANS Institute - Marcus Sachs, the SANS Institute - Chris Brenton - Josh Wright, the SANS Institute - Thanks to the SANS incident handlers for additional notes and assistance. - Thanks also to the authors of the above web resources.
(Version 0.3.2, 4/20/2004)
========
Call For Experts, SANS Top 20 List for 2004 SANS is looking for experts to participate in the development of the 2004 list of Top 20 Vulnerabilities. If you are interested in helping drive and develop the 2004 research please contact the Editor, Mr. Ross Patel (rpatel@sans.org), with the following details: your name organisation you represent contact details (inc. email and phone) a very brief synopsis of your sphere of specialism
=========
Marcus H. Sachs The SANS Institute
-- MoveOn.org -MFSO.org -ArnoldWatch.org - DigitalConsumer.org - FTCR.org - Privacy.org - Adbusters.org - Eff.com - Democraticmedia.org - HealthPrivacy.org - Hacktivismo.com - ClearChannelSucks.org - Epic.org |
|
  bewale Killemall Premium join:2000-08-08 Royal Oak, MI clubs: 1 edit | I'm surprised no one is mentioning IPv6 as possibly getting a beneficial push for deployment as a result of this. -- "In a world of compromise, some don't." |
|
 Daemon Premium join:2003-06-29 San Francisco, CA | The flaw is in TCP. Since TCP over IPv4 is the same as TCP over IPv6, it wouldn't make much of a difference.... -- -Ryan Find me in the networking and Microsoft help forums |
|
 hescominsoon
join:2003-02-18 Brunswick, MD | interesting not a peep on inet-access, nanog, full-disclosure, or vulnwtch over this one.. -- God Blesshttp://www.emmanuelcomputerconsulting.com |
|
  Steve I'm a PC, so shut up Consultant join:2001-03-10 Yorba Linda, CA | Ya heard it here first  |
|
  EGeezer Go Bobcats Premium join:2002-08-04 Country!
·Callcentric
·RoadRunner Cable
·AT&T CallVantage
3 edits | reply to Steve More links on the TCP IP Vulnerability
FYI, just a few I thought were informative;
TechRepublic's story has several good hyperlinks to other resources
»ct.com.com/click?q=b5-50ycQ7UYvH···5c5UWutC
Cert chimes in - I believe this link will be updated soon.
»www.us-cert.gov/cas/techalerts/T···11A.html
Washington Post article (registration needed to view but my experience is they don't spam) »www.washingtonpost.com/wp-dyn/ar···r20.html
A nice brief high level technical overview at X-Force (ISS)
»xforce.iss.net/xforce/alerts/id/170
Finally, The UK advisory - If this is a dup post, my apologies - »www.uniras.gov.uk/vuls/2004/236929/index.htm
I notice terrorist.net, Paul Watson's site, is experiencing high response times... Must be popular - I didn't see anything particularly interesting there - yet -
Edit - Following is a paraphrase of what an associate sent me on the matter. It provides a bit of insight from an extremely knowledgable viewpoint. Thanks to Matt Curtin of Interhack for his observations! ********* see »www.ietf.org/internet-drafts/dra···e-00.txt
Attacks with spoofed RST and SYN packets were documented in Bob Morris's 1985 paper, A weakness in the 4.2BSD UNIX TCP/IP software. That led to changes, most notably in the randomness of TCP sequence numbers. They were sequential in most implementations.[1]
That they can be done blindly is apparent and brute-force attacks are possible. There is nothing new new here, except the explicit statement that BGP is vulnerable.
Anything that sits atop TCP is vulnerable. But now, all of the news is talking about this like the sky is falling. This is a huge problem, but has been publicly known for years. We've been encouraging people to use IPSEC wherever possible for exactly this kind of thing.
Footnotes: [1] »citeseer.ist.psu.edu/context/220838/0
******End of paraphrase********
Edit - added ietf June 2003 BGP vulnerability analysis »www.ietf.org/internet-drafts/dra···n-00.txt
EG -- Those who cannot remember the past - must have had quite a party. |
|
  bewale Killemall Premium join:2000-08-08 Royal Oak, MI clubs:
1 edit | reply to Daemon Re: The Sky Is Falling Again...
said by Daemon : The flaw is in TCP. Since TCP over IPv4 is the same as TCP over IPv6, it wouldn't make much of a difference....
I disagree. IPv6 uses crypto at layer 3 to authenticate both ends of the conversation. The only attack vector of the TCP vulnerability announced by CERT is a "man in the middle attack", and hence would fail under IPV6 since v6 would drop the forged "man in the middle" packets at layer 3 - not even making it to the TCP layer's vulnerability (layer 4).
The fact that something like this isn't pushing IPv6's implementation tells me IPv6 is all but dead. Get the shovel out. -- "In a world of compromise, some don't." |
|
 Microsoft 98
join:2003-01-22 Micmac, NS
1 edit | reply to Steve Re: "Core Internet technology found vulnerable"
Alright what's the fix for this my Dlink 604 happen to reset itself just few minutes ago and yesterday.
I disagree I'm just a regular cable home user with 2 computers under dlink 604 with windows 98se on both pc's and my router happen to reboot itself twice so far and I got the current firmware patch which is now old Current Firmware Version: 2.20 -- http ://www.freewebs.com/welcome_to_my_store |
|
  Steve I'm a PC, so shut up Consultant join:2001-03-10 Yorba Linda, CA | Personal home routers are not likely to be vulnerable, as (among other reasons) they don't run BGP.
Steve |
|
 Microsoft 98
join:2003-01-22 Micmac, NS | reply to Steve Is that a fact so far dude? I never had my router reboot itself before hmmmmmm -- http ://www.freewebs.com/welcome_to_my_store |
|