OS-Specific Security Issues

Modified 10 March 2008

Cisco Router Security

Here are some great references on current best practices for Cisco router configurations:

• http://www.cymru.com/~robt/Docs/Articles/secure-ios-template.html
• http://www.cymru.com/~robt/Docs/Articles/secure-bgp-template.html
• http://www.itprc.com/security.htm
• http://www.securite.org/presentations/secip/
• Center for Internet Security benchmarks: http://www.cisecurity.org/

Linux

• My overview of how to harden a default Linux or OpenBSD installation
• My overview of how to set up SSH
• UNIX IP Stack Tuning Guide — http://secinf.net/info/unix/ip-stack-tuning.html
• Securing and optimizing Linux — there's a great book available for free download at http://www.packetstormsecurity.com/papers/unix/ on configuring and tuning Linux for improved security and performance. It's only one of several great references available there.
• Libsafe — protect against buffer overflow and "stack smashing". http://www.research.avayalabs.com/project/libsafe/
• FormatGuard — Enhanced version of glibc libraries: http://www.wirex.com/Immunix/downloads.html or http://www.ibiblio.org/pub/linux/distributions/immunix/
• Saint Jude Linux Kernel Module — discover and block attempts to gain inappropriate privileges:
  • http://www.sourceforge.net/projects/stjude/
  • http://www.packetstormsecurity.com/linux/security/
• Linux Kernel Patches
  • BufOverA — Linux kernel patches to protect against buffer overflow: http://www.iac.rm.cnr.it/newweb/tecno/software/indexsoftware.htm
  • LIDS — intrusion detection system patch for the Linux kernel, protect files, raw devices, processes: http://www.packetstormsecurity.com/linux/security/
  • Secure-Linux kernel patches — prevent cookbook buffer-overflow attacks: http://www.openwall.com/linux/
  • Stealth Linux kernel patches — prevent IP stack fingerprinting:
    • http://www.innu.org/~sean/
    • http://www.energymech.net/madcamel/fm
  • Medusa DS9 — kernel patches to monitor syscalls, filesystem actions, and processes, plus a security daemon: http://medusa.fornax.sk/
• Mason — automated Linux firewall rule-building tool: http://mason.stearns.org/
• Center for Internet Security benchmarks: http://www.cisecurity.org/
• Linux-based firewalls:
  • Astaro Security Linux — http://www.astaro.com/
  • Check Point Express — also available for Solaris and Windows — http://www.checkpoint.com/
  • Sun iForce VPN/Firewall — yes, it runs a hardened version of Linux, not Solaris — http://www.sun.com/
  • Trustix Enterprise Firewall — http://www.trustix.com/
• How to integrate any sort of *NIX into a Kerberos realm based on a Windows Active Directory server (ugh): http://www.cromwell-intl.com/unix/kerberos.html

Solaris Security

• UNIX IP Stack Tuning Guide — http://secinf.net/info/unix/ip-stack-tuning.html
• My overview of how to set up SSH
• http://secinf.net/info/unix/solaris_hardening.html
• http://secinf.net/info/unix/securing.html
• http://secinf.net/info/unix/harden1.html
• http://secinf.net/info/unix/basic_security.html
• http://secinf.net/info/unix/security-faq.html
• Center for Internet Security benchmarks: http://www.cisecurity.org/
• Hardening Solaris user authentication:
  • If you have Solaris 8 or earlier, you are stuck with the mid-1970's "classic UNIX" authentication. Only the first 8 password characters matter, and they're hashed in a way that was fairly secure against brute-force attack by, oh, common computers of the mid-1970's...
  • If you have Solaris 10 or later, it uses much more secure password hashing by default. Arbitrarily long pass phrases, hashed with MD5.
  • If, on the other hand, you have Solaris 9, while it is capable of the stronger method, you have to specify it.
    • First, modify /etc/default/passwd and change the line now reading:
      PASSLENGTH=6
      to a more ambitious number.
    • Edit /etc/security/policy.conf, first verifying that it contains this line:
      CRYPT_ALGORITHMS_ALLOW=1,2a,md5
      
    • Now un-comment this line by removing the leading "#":
      CRYPT_ALGORITHMS_DEPRECATE=__UNIX__
      
    • Change the default hashing algorithm like this:
      CRYPT_DEFAULT=1
      
    • Once you've done that, you will need to re-set the passwords with passwd to put the hashes into the more secure form.
    • The algorithms are as follows, note that the "1/2a/md5" notation was taken from Cisco:
      • 1 — BSD and Linux (GNU) style MD5 hashing. This is probably the best for interoperability, and should be quite strong.
      • 2a — OpenBSD style hashing, using a 1024-round lossy variation of Blowfish as a one-way operation. It has a password length limit, see the manual pages for crypt.conf, crypt_bsdmd5, and crypt_bsdbf.
      • md5 — Some Solaris-specific MD5 hashing.
• How to integrate any sort of *NIX into a Kerberos realm based on a Windows Active Directory server (ugh): http://www.cromwell-intl.com/unix/kerberos.html

ULTRIX/ OSF/1 / Digital Unix / Tru64 Unix Security

The BSDish variant of Unix from Digital/Compaq/HP/whatever has bugs and fixes: http://www.insecure.org/sploits_ultrix.html

HP-UX Security

HP-UX bugs and fixes: http://secinf.net/info/unix/secureHP-UX.html

IRIX Security

Silicon Graphic's variant of Unix has bugs and fixes: http://www.insecure.org/sploits_irix.html

AIX Security

IBM's variant of Unix has bugs and fixes: http://www.insecure.org/sploits_aix.html

DOS Security

Go to the COAST archive: ftp://coast.cs.purdue.edu/pub/tools/dos

AS/400 Security

• See the PSSecure and PSAudit products from PentaSafe. http://www.pentasafe.com/
• Also see the StandGuard and StandGuardAudit products from Bytware. http://www.bytware.com/
• There was an interesting article about Bytware AS/400 security products at Midrange Server magazine.

Novell Security

• Security tools — ftp://coast.cs.purdue.edu/pub/tools/novell/
• Novell security risks — http://www.nmrc.org/
• Details of how to easily defeat Novell's encrypted remote password, giving administrative access to your Novell servers — http://oliver.efri.hr/~crv/security/bugs/Others/nware12.html

VMS Security

• Axent has VAX/VMS security tools. http://www.axent.com/
• Think you're secure because you run VMS, not Unix? See the first chapter of UNDERGROUND — Tales of Hacking, madness and obsession on the Electronic Frontier (ISBN 1-86330-595-5), http://www.underground-book.com

Windows Security Abandon hope, all ye who enter here

Jim Allchin was a vice-president and later co-president of Microsoft. He retired from Microsoft as of 30 January 2007, the day on which Microsoft officially released their Windows Vista operating system to consumers. Allchin was co-president of Microsoft's Platforms & Services Division, was the manager of the Vista project, and led the development of a number of Microsoft's operating systems. Allchin provides some interesting information:

• "I am not sure how the company lost sight of what matters to our customers (both business and home) the most, but in my view we lost our way. I think our teams lost sight of what bug-free means, what resilience means, what full scenarios mean, what security means, what performance means, how important current applications are, and really understanding what the most important problems [our] customers face are. I see lots of random features and some great vision, but that doesn't translate into great products.
  I would buy a Mac today if I was not working at Microsoft. ... Apple did not lose their way. ..."
  • The above is from the Allchin memo to Bill Gates and Steve Ballmer, January 2004
  • http://en.wikipedia.org/wiki/James_Allchin
  • http://www.iowaconsumercase.org/010807/PLEX_7264.pdf
• Then there is the problem of privilege escalation through the Win32 API. Microsoft says this is not fixable but not really a problem, which is true unless you really don't want people breaking your OS. While testifying before the U.S. Department of Justice in an anti-trust case, Allchin referred to this as a fundamentally unsecurable design representing a threat to U.S. national security due to the U.S. government's reliance on Windows. Allchin then mentioned the Windows message-queuing subsystem, which allows for what's known as the "Shatter" attack.
  • http://security.tombom.co.uk/shatter.html
  • http://www.governmentsecurity.org/forum/index.php?showtopic=8989
  This was at least partly fixed in Vista, but some issues remain. Oh, and after Microsoft said "Outsiders cannot be allowed to see the source code as that would damage US national security", they allowed the government of the People's Republic of China to view the source code. See, the PRC said they didn't want to buy Windows unless they could see the source code, and Microsoft didn't want to miss out on a big sale....

It appears that the guy in charge of the Microsoft operating systems has very little confidence in them. Why should we contradict him?

Then there are the three highly placed Microsoft executives whose internal communications were brought to light in a US District Court case, described in a New York Times article 9 March 2008. Their frustrations were caused by a lack of functionality and support in Vista, which at the time of their problems had been released as a supposedly finished product and was being sold for full retail price. The angry executives included:

• Jon A Shirley, a Microsoft board member, former president, and chief operating officer, "upgraded" two XP machines to Vista. Then he discovered that his printer, scanner, and film scanner all lacked Vista drivers. He had to go back to XP on one machine just to continue using those peripherals.
• Mike Nash, a Microsoft vice president who oversees Windows product management, bought a laptop with a "Windows Vista Capable" logo. That laptop lacked the needed graphics chip, could not run his favorite video-editing software, and could only run "a hobbled version of Vista". "I got burned", he said, "I now have a $2,100 e-mail machine."
• Steven Sinofsky, Microsoft's senior vice president responsible for Windows, heard about Jon Shirley's problems and said that drivers are missing in every category in Vista — "This is the same across the whole ecosystem."

If you want to reduce your security risk due to Windows:

1. Use any other operating system. Really. Most users need little more than a web browser, an e-mail tool, and something to handle documents.
2. If you really must use Windows on some systems, then do not use Explorer for anything. Beyond profound code quality issues, aspects of its design are fundamentally insecure and unsecureable. Use any other browser, most people find the Mozilla Firefox browser an excellent tool. Most people also find that getting rid of Exploder means, for the most part, an end to spyware, and many phishing attacks become more obvious.
3. Given that, do not use Outlook for anything, as it silently uses some of the most insecure components of Exploder and the user can't prevent that. Mozilla Firefox comes with Thunderbird, an integrated e-mail client.
4. Use the NTFS file system, but don't expect it to protect you from booting off a Knoppix CD.
5. If you use Kerberos, rip out Microsoft's weakened version and use real Kerberos, available for free from MIT. Weakened? Yes, their silly "pre-authentication" violation of the Kerberos rules supports a known-plaintext attack by an attacker.
6. Finally, try to use any other operating system in place of Windows.

On to the Windows security list. Remember that "NT" is Microsoft's term for an entire family of operating systems, including for NT 3.x, NT 4.x, Windows 2000, Windows XP, Windows 2003, and Windows Vista.

• A good discussion of Windows rootkits — gets you into the technical details of how a rootkit works: http://www.securityfocus.com/infocus/1850
• Breaking in with bootable media:
  • The Windows "rescue" diskette, a.k.a. the NT break-in diskette: http://home.eunet.no/~pnordahl/ntpasswd
  • Or, you could do similar things with the Knoppix Linux live DVD: http://www.knoppix.org/
• NSA recommendations for securing Windows: http://www.nsa.gov/winsecurity/win2k/download.htm
• Disturbing facts about Windows file system insecurity
  • The newer members of the Windows NT family support EFS, Encrypting File System, which originally seemed like a decent design:
    • Every file is encrypted using a symmetric cipher and a key randomly generated when the file was created.
    • The collection of keys is stored in a file encrypted with a key based on the user's password.
    • You type your password to login, and so your login session has access to the key store.
    • If you change your password while logged in, your old (current) password is used to decrypt the key store and then the new (future) password is used to encrypt the key store.
    • The remaining problem is that if you use EFS and you reset a users's password, you lose the ability to decrypt that key store and thus all of that user's files are lost as undecipherable ciphertext.
    • The good news (in this very specific case) is that Windows password security is so weak that you can use something like Rainbow Tables to break the user's password in minutes instead of resetting it and losing the data.
  • The disturbing part is this disclosure on one of Microsoft's pages:
    Windows NT zero-fills memory and zeroes the sectors on disk where a file is placed before it presents either type of resource to an application. owever, object reuse does not dictate that the space that a file occupies before it is deleted be zeroed. This is because Windows NT/2K is designed with the assumption that the operating system controls access to system resources. However, when the operating system is not active it is possible to use raw disk editors and recovery tools to view and recover data that the operating system has deallocated. Even when you encrypt files with Win2K's Encrypting File System (EFS), a file's original unencrypted file data is left on the disk after a new encrypted version of the file is created.
    
  • So.... EFS only matters if you agree to play by Microsoft's contrived rules and only use Microsoft Windows to operate all computers at all times. But boot from Knoppix and there is cleartext data lying all over the place. This is a strangely narrow meaning of the term "Encrypting File System"!
  • See the page describing this strangeness at: http://www.microsoft.com/technet/sysinternals/FileAndDisk/SDelete.mspx
• An article to frighten you about Microsoft design trends: http://www.hevanet.com/peace/microsoft.htm
• Center for Internet Security benchmarks: http://www.cisecurity.org/
• http://www.ntsecurity.net/
• http://www.ntbugtraq.com/
• http://www.l0pht.com/advisories.html
• http://www.warzone.org/
• A great collection of Windows exploits: http://www.dhp.com/~fyodor/sploits.html
• BackOriface, NetBus, and other attacks detailed: http://www.nttoolbox.com/ —
• Microsoft exploit details: http://www.insecure.org/sploits_microshit.html
• The NetBus attack detailed: http://www.nwi.net/~pchelp/nb/netbus.htm/
• http://ntshop.net
• http://sysinternals.com
• Join the NT security mailing list, at request-ntsecurity@iss.net
• Microsoft's implementation of PPTP, on which they base their VPN's, is seriously broken (although they claim it isn't):
  • http://www.counterpane.com/pptp.html
  • http://www.geek-girl.com/bugtraq/1999_1/0664.html
  • http://oliver.efri.hr/~crv/security/bugs/NT/pptp2.html
• The classic John Kirch paper describes in detail why various Unix operating systems out-perform the Windows NT family ("NT" is Microsoft's term for NT 3.x, NT 4.x, Windows 2000, Windows XP, Windows 2003, and Windows Vista): http://www.lege.com/unix-nt/
• Windows NT Security Tools
  • Kane Security Monitor (network traffic and log analysis), http://www.intrusion.com
  • RealSecure 1.0 (traffic analysis and attack signature detection), Internet Security Systems, http://www.iss.net
  • NukeNabber scans and logs incoming TCP/IP traffic, including packet attacks like Boink, TearDrop, Land, etc. http://www.dynamsol.com/
  • There are some tools to clean BackOriface, Netbus, and other attack code from infected NT machines.
    • http://www.pandasoftware.com
    • http://www.privsoft.com — see BOClean32
    • http://www.dynamsol.com/puppet — see Puppet
• For Windows security problems in detail, see Byte, November 1997, pp 81-86 for details of how Microsoft re-invents broken wheels. Many problems that were discovered, and fixed, in other operating systems over the past two to twenty years are being repeated by Microsoft! Also see Computers and Security, vol 17, no 2, pp 100-106 for details.
• See my page on user authentication for details on just how weak the default Windows design is.
• Internet Explorer bugs: http://www.tbtf.com/resource/ms-sec-exploits.html
• How Microsoft's poor (missing?) protection of the master cryptographic key exposes all of a user's net communication: http://www.cs.auckland.ac.nz/~pgut001/pubs/breakms.txt and http://www.techweb.com/wire/story/TWB19980123S0007

Comparing Windows to Linux and BSD

What is wrong with Linux?

It is as disorganized and resistant to organization as a herd of cats.

The kernel of the operating system itself is fairly good. But a Linux distribution is largely a pile of things that aren't the kernel, and which tend to lower security:

• Application software — possibly buggy but relied on and assumed secure by users.
• Graphical environments — even worse! More likely to be buggy due to the volume of code, and just like Windows they hide details from users and automatically invoke programs in response to mouse events.
• Distribution installation and configuration tools — the installation process makes some security related assumptions on behalf of the administrator. The configuration tools used for ongoing administration may not be as bad, but they still make some assumptions that the administrator will not notice without great care.

Also, Linux is not very good for playing games. Apparently this matters a lot more than security to an awful lot of people.

What is wrong with BSD?

The same problems as Linux, just to far less degree.

While there are just three BSD implementations — FreeBSD, NetBSD, and OpenBSD, and their releases are far better controlled those for Linux, the BSDs use the same application software and graphical environment (GNU, KDE, Gnome, et al) as Linux. While their installation tools assume quite a bit of *NIX expertise (you'd better know how to set up a BSD partition/slice scheme by hand!), they may still hide some details from the installer.

BSD is not very good for playing games, either, if you care more about that than security.

What is wrong with Windows?

Three crucial components seem to be far more intertwined in Windows than in other operating systems:

• The operating system
• The graphical interface
• The environment of user's processes

The accepted method for administering the system is to login to the graphical interface as Administrator and use graphical tools. There is next to no separation of privileges.

Compare that to the UNIX model where the operating system and the graphical interface are relatively separate, and where cautious administrators log in as unpriviliged users. Only to the extent absolutely necessary do they elevate privileges, using su or Role-Based Access Control (RBAC) tools.

Then there are the other really bad design decisions — the window message queueing API, really questionable TCP/IP design decisions (like file and print sharing over IP broadcast rather than IP multicast!), etc.

As for some commentary on the Windows source code that leaked in early 2004, see: http://www.kuro5hin.org/story/2004/2/15/71552/7795

Finally, don't forget hubris, which has caused trouble for its practitioners since the Illiad and Odyssey. Microsoft's continued claims that their latest expensive product is far more secure or reliable will just invite attacks.

However, Windows is very good for playing games if that's what matters to you.

Security Page

© Bob Cromwell Jul 2008. Created with /bin/vi, hosted on OpenBSD with Apache.    Root password available here