Cisco, haven't we learned anything? (technician reset)
In this (http://blogs.securiteam.com/wp-admin/post.php?action=edit&post=207) recent Cisco advisory, the company alerts us to a security problem with Cisco MARS (Cisco Security Monitoring Analysis and Response System). The security issue is basically a user account on the system that will give you root when accessed. The account is: 1. Hidden. 2. Default. 3. With a pre-set password. In other words, this is a journey back 10 years when technicians would commonly have special keys (actual keys, electronics or passwords) to access a device if they have to troubleshoot it for anything, or say? the user lost his password. People used to trade these keys online and hidden accounts were a thing of common practice. Today people still trade commonly used default passwords but it is not as popular as it used to be, at least in the online world. On the other hand, the most common practice to hack routers today, is still to try and access the devices with the notoriously famous default login/password for Cisco devices: cisco/cisco. Cisco/cisco is the single most used default password of our time. It got more routers pwned than any exploit in history, and it still does. One would think that a company such as Cisco, especially with this history, would stay away from such ?default? accounts? but the fact that this account is hidden makes it something different. It makes it a backdoor. One much like those used by the Bad Guys. Now? if Cisco knowingly put it there, shame on them. If somebody put it there without their knowledge? well, shame on them. This is indeed a vulnerability, as in a weakness. It is not however a software coding bug that may result in say? a buffer overflow. It is a part of the design of the system. Cisco disclosing this is very nice and commendable, but perhaps they should also let us know whether this was indeed a backdoor somebody put in their system or if it was part of the design? I love eastereggs. I just don?t like surprises in system privileges or backdoors, especially not in a security monitoring and response product. I very much doubt it was anything else but a part of the design but that should be admitted to. As the advisory states: "No other Cisco products are currently known to be affected by this vulnerability." Okay, but how about other vulnerabilities of this type? Are there any more backdoors to other Cisco products? If not, why wouldn?t they just come out and say that? ?There are NO other such backdoors in our products?. I?d even be happy with: ?To our knowledge, there are no other vulnerabilities of this type in our products.? This is not a bug. One can never be sure ALL bugs are eliminated ? however hard one may try. One CAN admit to having no such features in other products, though. Once again we fall upon re-naming of a feature as a bug or a bug as a feature to make the problem sound less severe. IN this case, the judgement is plain and simple: If Cisco were Bad Guys, this is a backdoor. As Cisco are Good Guys, this is a technician reset. Terminology? What?s the difference? The difference is that Cisco are not Bad Guys. If they disclosure a problem they should do it fully, because as a client, I am now concerned. This reminds me of Ciscogate but not for obvious reasons. That was a bad event for everybody involved. It reminds me of the very issue Mike Lynn discussed: Remote exploitation for Cisco is possible, while so far Cisco disclosed all these problems as DoS vulnerabilities. I am not saying Cisco did that on purpose, but in THIS case they CAN set my mind at ease. Why don?t they? Gadi.
On Thu, 12 Jan 2006, Gadi Evron wrote:
In this (http://blogs.securiteam.com/wp-admin/post.php?action=edit&post=207) recent Cisco advisory, the company alerts us to a security problem with Cisco MARS (Cisco Security Monitoring Analysis and Response System).
The security issue is basically a user account on the system that will give you root when accessed. ... Now? if Cisco knowingly put it there, shame on them. If somebody put it there without their knowledge? well, shame on them.
Cisco acquired Protego in Dec 2004 and thereby acquired MARS: http://www.infoworld.com/article/04/12/20/HNciscoprotego_1.html Cisco didn't put it in there - they bought the bug for $65M. :-)
Okay, but how about other vulnerabilities of this type? Are there any more backdoors to other Cisco products? If not, why wouldn?t they just come out and say that? ?There are NO other such backdoors in our products?.
I am sure there are more. The previous one I remember was with their Riverhead purchase: http://www.cisco.com/en/US/products/products_security_advisory09186a008037d0... and before that was: http://www.cisco.com/en/US/products/products_security_advisory09186a00802119... but I don't know which company was purchased to introduce that one. I think Cisco just doesn't check the product closely enough and trusts the R&D coders and doesn't introduce an external security QA to the product being purchased. -Hank
Hi, NANOGers. ] On the other hand, the most common practice to hack routers today, is ] still to try and access the devices with the notoriously famous default ] login/password for Cisco devices: cisco/cisco. This is NOT a default password in the IOS. The use of "cisco" as the access and enable passwords is a common practice by users, but it isn't bundled in the IOS. I've heard it began in training classes, where students were taught to use "cisco" as the passwords. Oh, and for those of you who think it mad leet to use "c1sc0" as your access and enable passwords, the miscreants are on to that as well. ;) We've seen large, massively peered and backbone routers owned through this same technique. We've even seen folks who have switched to Juniper, yet continue to use "cisco" as the login and password. :( The nice thing about cooking up blame is that there is always enough to serve everyone. Thanks, Rob. -- Rob Thomas Team Cymru http://www.cymru.com/ ASSERT(coffee != empty);
Just as an offshoot discussion, what's the state-of-the-art for AAA services? We use an modified tacacs server for multi-factor authentication, and are moving towards a model that supports single-use/rapid expiration passwords, with strict control over when and how local/emergency authentication can be used. I'd be interested in that discussion, on or offlist. - billn On Thu, 12 Jan 2006, Rob Thomas wrote:
Hi, NANOGers.
] On the other hand, the most common practice to hack routers today, is ] still to try and access the devices with the notoriously famous default ] login/password for Cisco devices: cisco/cisco.
This is NOT a default password in the IOS. The use of "cisco" as the access and enable passwords is a common practice by users, but it isn't bundled in the IOS. I've heard it began in training classes, where students were taught to use "cisco" as the passwords.
Oh, and for those of you who think it mad leet to use "c1sc0" as your access and enable passwords, the miscreants are on to that as well. ;)
We've seen large, massively peered and backbone routers owned through this same technique. We've even seen folks who have switched to Juniper, yet continue to use "cisco" as the login and password. :(
The nice thing about cooking up blame is that there is always enough to serve everyone.
Thanks, Rob. -- Rob Thomas Team Cymru http://www.cymru.com/ ASSERT(coffee != empty);
I've been pretty happy with Cisco ACS - fairly solid, good reporting, once set up it seems to Just Work. John On Thu, Jan 12, 2006 at 11:00:10AM -0800, Bill Nash wrote:
Just as an offshoot discussion, what's the state-of-the-art for AAA services? We use an modified tacacs server for multi-factor authentication, and are moving towards a model that supports single-use/rapid expiration passwords, with strict control over when and how local/emergency authentication can be used.
I'd be interested in that discussion, on or offlist.
- billn
On Thu, 12 Jan 2006, Rob Thomas wrote:
Hi, NANOGers.
] On the other hand, the most common practice to hack routers today, is ] still to try and access the devices with the notoriously famous default ] login/password for Cisco devices: cisco/cisco.
This is NOT a default password in the IOS. The use of "cisco" as the access and enable passwords is a common practice by users, but it isn't bundled in the IOS. I've heard it began in training classes, where students were taught to use "cisco" as the passwords.
Oh, and for those of you who think it mad leet to use "c1sc0" as your access and enable passwords, the miscreants are on to that as well. ;)
We've seen large, massively peered and backbone routers owned through this same technique. We've even seen folks who have switched to Juniper, yet continue to use "cisco" as the login and password. :(
The nice thing about cooking up blame is that there is always enough to serve everyone.
Thanks, Rob. -- Rob Thomas Team Cymru http://www.cymru.com/ ASSERT(coffee != empty);
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Yo Rob! On Thu, 12 Jan 2006, Rob Thomas wrote:
This is NOT a default password in the IOS.
Uh, wrong. Check out the doc for the Cisco AIR-AP1220. Ver 12.01T1 RGDS GARY - --------------------------------------------------------------------------- Gary E. Miller Rellim 20340 Empire Blvd, Suite E-3, Bend, OR 97701 gem@rellim.com Tel:+1(541)382-8588 Fax: +1(541)382-8676 -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.1 (GNU/Linux) iD8DBQFDxrGB8KZibdeR3qURAvHaAJ0Q7Lt6rkj+jQNqnsMdpJX+DfldywCgidbI qX+Hm/+94o/W4F5sN6NRuzU= =EfCd -----END PGP SIGNATURE-----
Rob Thomas wrote:
Hi, NANOGers.
] On the other hand, the most common practice to hack routers today, is ] still to try and access the devices with the notoriously famous default ] login/password for Cisco devices: cisco/cisco.
This is NOT a default password in the IOS. The use of "cisco" as the access and enable passwords is a common practice by users, but it isn't bundled in the IOS. I've heard it began in training classes, where students were taught to use "cisco" as the passwords.
Actually, and fairly recently, this IS a default password in IOS. New out-of-box 28xx series routers have cisco/cisco installed as the default password with privilege 15 (full access). This is a recent development. To be fair, the box also has a huge default login banner urging the user to delete that username/password pair. But we all know how much attention is paid to huge, verbose banners, disclaimers, click-to-agree dialog boxes, etc. -- Jay Hennigan - CCIE #7880 - Network Administration - jay@west.net NetLojix Communications, Inc. - http://www.netlojix.com/ WestNet: Connecting you to the planet. 805 884-6323
Actually, and fairly recently, this IS a default password in IOS. New out-of-box 28xx series routers have cisco/cisco installed as the default password with privilege 15 (full access). This is a recent development.
This is hardly only cisco's problem. Most office routers I've dealt with also come with default username/password and on occasions when I dealt with existing installation those passwords have rarely been changed. What should really be done (BCP for manufactures ???) is have default password based on unit's serial number. Since most routers provide this information (i.e. its preset on the chip's eprom) I don't understand why its so hard to just create simple function as part of software to use this data if the password is not otherwise set. -- William Leibzon Elan Networks william@elan.net
william(at)elan.net wrote:
Actually, and fairly recently, this IS a default password in IOS. New out-of-box 28xx series routers have cisco/cisco installed as the default password with privilege 15 (full access). This is a recent development.
This is hardly only cisco's problem. Most office routers I've dealt with also come with default username/password and on occasions when I dealt with existing installation those passwords have rarely been changed.
True. However I much prefer the old way that Cisco did it. No default passwords on the box at all. But, no remote administration at all until a password was set on the console. Now, there is a default cisco/cisco. Newbie admin creates a new user/pass, tests thinks it's secure, fails to remove the default, game over.
What should really be done (BCP for manufactures ???) is have default password based on unit's serial number. Since most routers provide this information (i.e. its preset on the chip's eprom) I don't understand why its so hard to just create simple function as part of software to use this data if the password is not otherwise set.
The old-school Cisco way works for me. Default is no password if you have physical access, but no remote access. -- Jay Hennigan - CCIE #7880 - Network Administration - jay@west.net NetLojix Communications, Inc. - http://www.netlojix.com/ WestNet: Connecting you to the planet. 805 884-6323
On Thu, 12 Jan 2006, Jay Hennigan wrote:
What should really be done (BCP for manufactures ???) is have default password based on unit's serial number. Since most routers provide this information (i.e. its preset on the chip's eprom) I don't understand why its so hard to just create simple function as part of software to use this data if the password is not otherwise set.
The old-school Cisco way works for me. Default is no password if you have physical access, but no remote access.
That works too and is most secure way. But its often enough that small offices would not have person who can fix the system and its not always possible to get network guy to come in right a way. It is good for those cases to be able to ask somebody onsite to just look at the back and dictate the serial# by phone. -- William Leibzon Elan Networks william@elan.net
Actually, and fairly recently, this IS a default password in IOS. New out-of-box 28xx series routers have cisco/cisco installed as the default password with privilege 15 (full access). This is a recent development.
This is hardly only cisco's problem. Most office routers I've dealt with also come with default username/password and on occasions when I dealt with existing installation those passwords have rarely been changed.
What should really be done (BCP for manufactures ???) is have default password based on unit's serial number. Since most routers provide this information (i.e. its preset on the chip's eprom) I don't understand why its so hard to just create simple function as part of software to use this data if the password is not otherwise set.
Ex: Thot's how a Netscreen 5 works after a reset. The password is the serial # if I remember correctly. -M<
In message <200601130141.k0D1fiZ1007762@world.std.com>, Martin Hannigan writes:
Actually, and fairly recently, this IS a default password in IOS. New out-of-box 28xx series routers have cisco/cisco installed as the default password with privilege 15 (full access). This is a recent development.
This is hardly only cisco's problem. Most office routers I've dealt with also come with default username/password and on occasions when I dealt with existing installation those passwords have rarely been changed.
What should really be done (BCP for manufactures ???) is have default password based on unit's serial number. Since most routers provide this information (i.e. its preset on the chip's eprom) I don't understand why its so hard to just create simple function as part of software to use this data if the password is not otherwise set.
Ex: Thot's how a Netscreen 5 works after a reset. The password is the serial # if I remember correctly.
How much entropy is there in a such a serial number? Little enough that it can be brute-forced by someone who knows the pattern? Using some function of the serial number and a vendor-known secret key is better -- until, of course, that "secret" leaks. (Anyone remember how telephone credit card number verification worked before they could do full real-time validation? The Phone Company took a 10-digit phone number and calculated four extra digits, based on that year's secret. Guess how well that secret was kept....) --Steven M. Bellovin, http://www.cs.columbia.edu/~smb
On Thu, 2006-01-12 at 21:05:52 -0500, Steven M. Bellovin proclaimed...
How much entropy is there in a such a serial number? Little enough that it can be brute-forced by someone who knows the pattern? Using some function of the serial number and a vendor-known secret key is better -- until, of course, that "secret" leaks. (Anyone remember how telephone credit card number verification worked before they could do full real-time validation? The Phone Company took a 10-digit phone number and calculated four extra digits, based on that year's secret. Guess how well that secret was kept....)
Hi Steven, I believe the Netscreen default password of a serial number can only be entered over the console (and possibly modem/aux) port(s).
On Thu, 2006-01-12 at 21:05:52 -0500, Steven M. Bellovin proclaimed...
How much entropy is there in a such a serial number? Little enough that it can be brute-forced by someone who knows the pattern? Using some function of the serial number and a vendor-known secret key is better -- until, of course, that "secret" leaks. (Anyone remember how telephone credit card number verification worked before they could do full real-time validation? The Phone Company took a 10-digit phone number and calculated four extra digits, based on that year's secret. Guess how well that secret was kept....)
Hi Steven,
I believe the Netscreen default password of a serial number can only be entered over the console (and possibly modem/aux) port(s).
Yes. Sorry, I left that out. -M<
In message <20060113023430.GI22251@catastrophe.net>, eric writes:
On Thu, 2006-01-12 at 21:05:52 -0500, Steven M. Bellovin proclaimed...
How much entropy is there in a such a serial number? Little enough that it can be brute-forced by someone who knows the pattern? Using some function of the serial number and a vendor-known secret key is better -- until, of course, that "secret" leaks. (Anyone remember how telephone credit card number verification worked before they could do full real-time validation? The Phone Company took a 10-digit phone number and calculated four extra digits, based on that year's secret. Guess how well that secret was kept....)
Hi Steven,
I believe the Netscreen default password of a serial number can only be entered over the console (and possibly modem/aux) port(s).
That works for me. But note William Leibzon's issue: That works too and is most secure way. But its often enough that small offices would not have person who can fix the system and its not always possible to get network guy to come in right a way. It is good for those cases to be able to ask somebody onsite to just look at the back and dictate the serial# by phone. If you have physical access, the root password matters a lot less (and if it's the serial number, the local attacker can just peer at the back). If you need secure remote access -- well, it's not easy with clueless local administrators. But there's much less excuse for clueless developers, like the ones who created the login/password pair that started this thread -- credentials that, according to one posting, are acceptable for remote access. --Steven M. Bellovin, http://www.cs.columbia.edu/~smb
This reminds me of Ciscogate but not for obvious reasons. That was a bad event for everybody involved. It reminds me of the very issue Mike Lynn discussed: Remote exploitation for Cisco is possible, while so far Cisco disclosed all these problems as DoS vulnerabilities. I am not saying Cisco did that on purpose, but in THIS case they CAN set my mind at ease.
Why don?t they?
I did not change my mind, but to be fair, I have to add: After writing this I’ve been made aware that this product was from a company Cisco bought not so long ago. This very same issue happened before (and more than once)... in one recent example with another company Cisco bought named Riverhead. It is true Cisco's PSIRT is one of the best to work with among vendors, even Mike Lynn said that Cisco PSIRT are some of the more decent people he worked with - "I've never had a problem with PSIRT". It is also true that Cisco can't find out about these until after they buy the companies, still, Cisco f*cked up, more than just once or twice, and we call it. This kind of a so-called "vulnerability" should not happen, or be disclosed, continually, in this particular fashion. Checking into new investments security-wise, especially with security products and external QA may help solve such issues in the future. Gadi.
On Fri, 2006-01-13 at 01:30:52 +0200, Gadi Evron proclaimed...
Checking into new investments security-wise, especially with security products and external QA may help solve such issues in the future.
Thank you for this interruption. We now returned to our scheduled programming, already in progress.
[ SNIP ]
It is true Cisco's PSIRT is one of the best to work with among vendors, even Mike Lynn said that Cisco PSIRT are some of the more decent people he worked with - "I've never had a problem with PSIRT".
PSIRT is great. After marketing and legal approval. This is why they can't possibly have "responsible" disclosure. :-) Ok, that's bait. -M<
participants (11)
-
Bill Nash -
eric -
Gadi Evron -
Gary E. Miller -
Hank Nussbacher -
Jay Hennigan -
John Kinsella -
Martin Hannigan -
Rob Thomas -
Steven M. Bellovin -
william(at)elan.net