As Evi Nemeth et al put it:
Unix was not designed with security in mind, and for that reason no UNIX system can be made truly secure.
While I think that statement is becoming less true with the compartmentalization now available via virtualization and with the increased security from using mandatory access control systems such as SELinux, it is still valid. Certainly the large number of successful compromises this semester shows that security is still a concern.
Even the most security-conscious of all of the mainstream Unixes, OpenBSD, has had its flaws.
The basic flaws are in Unix are
The classic formulation
Security = 1 / (1.072 * Convenience)
Protecting fed workers from themselves
By Austin Modine
July 30, 2009
US Congress wants to know if new federal laws are needed to protect government employees from accidental file-sharing.
A House of Representatives oversight committee gathered on Wednesday to discuss whether government workers getting their hands on peer-to-peer software poses a risk to privacy and national security.
At issue are numerous cases of federal government employees and contractors who install P2P software on computers without realizing the sensitive documents they expose for sharing. According to the committee chairman, this is a problem with the software rather than user.
...
Towns laid out several past cases of apparent accidental file-sharing that lead to major security breaches on LimeWire. In one, the social security numbers and family information for every master sergeant in the US Army was made available. Another security breach involving the Secret Service resulted in the leak of a file containing a safe house location for the First Family. ...
From the
http://www.theregister.co.uk/2009/07/30/house_committee_oversight_government_reform_p2p_security_hearing/
By Bruce Wallace, Times Staff Writer
March 21, TOKYO
So far it has spilled military secrets and the private phone numbers of TV stars, airport security access codes and elementary school children's grades.
And the dirty work of this computer virus may not be done.
With almost daily reports of more private information being pumped from personal computers and splashed over the Internet, there is a growing unease that Japan is under insidious attack from within.
The culprit is a digital worm that infects computers using the file-sharing Winny software, a Japanese computer program that, like the infamous Napster, was designed to allow people to easily swap music and movie files.
From the Los Angeles Times, March 21st, 2006 at
http://www.latimes.com/news/nationworld/world/la-fg-computer21mar21,0,5159274.story
Mystery malware and file sharing linked to third breach
By John Leyden
Published Wednesday 17th May 2006 16:06 GMT
Sensitive information about Japanese power plants has leaked online from a virus-infected computer for the second time in less than four months. Data regarding security arrangements at a thermoelectric power plant run by the Chubu Electric Power in Owase, Mie Prefecture in central Japan spilled online this week as a result of an unnamed virus infection, the Japan Times reports.
The name and addresses of security workers, along with other sensitive data including the location of key facilities and operation procedures, found its way onto file-sharing networks. A 40 year-old sub-contractor at the plant who installed the Share file sharing programs on his PC is suspected of provoking the security flap.
The power plant suffered a similar incident in January over data that found its way onto the Winny file sharing network, the most popular P2P network in Japan, which boasts an estimated 250,000 users. That incident provoked a management edict designed to prohibit the use of file sharing programs, so the occurrence of a similar problem only four months later is doubly embarrassing for Chubu Electric Power.
Chubu Electric is not the only power firm with problems in this area of net security, however. In June 2005, nuclear power plant secrets had been leaked from a PC belonging to an worker at Mitsubishi Electric Plant Engineering, anti-virus firm Sophos notes. That breach, just like the January security flap at Chubu Electric, was also linked to virus infection and the Winny file sharing program.
From The Register, May 17th, 2006 at
http://www.theregister.co.uk/2006/05/17/japan_power_plant_virus_leak/
The weakest link is usually the human element. Social engineering takes advantage of the fact that people generally are not distrustful, such as demonstrated by Nigerian 419 schemes and by phishing. Education is our best answer, and even then, education is only as good as the most recent attack — the latest scheme may catch even a user wary of previous methods.
Software bugs are the second major category for security compromises. Patching is the main defense as this category.
``Open doors'' are the third way. Some software still has backdoors built-in. It is often the case that people also leave the front door wide open. This is particularly true with items such as wireless routers, which are frequently configured ``to just work.'' As a system administrator responsible for other people's data, preserving both user access and confidentiality for that data, you need to be aware of keeping doors closed.
The most important concern is ensuring that users use appropriate passwords, even if the passwords are just one factor in a multi-factor environment. You can (1) suggest that they use good passwords (2) try to enforce that they use good passwords and (3) try to check that they are using good passwords (``John the Ripper'' and ``L0phtCrack'' continue to be kept up-to-date, unlike older programs such as ``Crack.'')
Password ageing: it is generally recommended that passwords be changed on a regular basis, although there has been back-and-forth discussion on that issue. Also, there is a contingent that are against any multi-use of a single password, with schemes such as OPIE that try to fix this problem.
Here's an example of the combination of weak passwords and bad configuration being exploited to create a bot-net:
Botnet of OpenWRT/DD-WRT devices``The people who bring you the DroneBL DNS Blacklist services, while investigating an ongoing DDoS incident, have discovered a botnet composed of exploited DSL modems and routers. OpenWRT/DD-WRT devices all appear to be vulnerable. What makes this worm impressive is the sophisticated nature of the bot, and the potential damage it can do not only to an unknowing end user, but to small businesses using non-commercial Internet connections, and to the unknowing public taking advantage of free Wi-Fi? services. The botnet is believed to have infected 100,000 hosts.''
From Slashdot, March 23, 2009 at
http://it.slashdot.org/article.pl?sid=09/03/23/2257252
In times previous, one notorious problem was that of processes that had to look into kernel memory to find information. This was particularly true of programs such as top and ps (today, as mentioned previously, we get around this problem by the /proc directory.) Devices that referred to kernel memory, such /dev/kmem often had inadequate and unsafe permissions, or programs such as top had too much privilege for their intended function.
Make sure that no files in /etc are publicly writable. There is no good reason for any file in that subdirectory to be writable to the public. No files in directories such as /usr/bin or /usr/lib (assuming /usr/unification) to be world writable.
Check your device files and make sure that important devices such as disk drives have appropriate permissions.
This is one area where the find program shines for one-off checks. While its syntax is recondite, it can help you discover all manner of interesting things.
find /etc -type f -perm +022 -ls # check /etc for any files that are
# group or world writable
find /etc -type d -perm +022 -ls # check /etc for any directories that are
# group or world writable
find /etc -type f -perm +6000 -ls # check /etc for any files that are setuid or
# setgid
find /etc -mtime -14 -ls # check to see what files in /etc/ have
# been modified recently
Remote logging is common in larger installations.
Even with standard syslog, this is very easy to set up using the ``@'' syntax. For instance, the syslog.conf man page gives these example lines (slightly modified for clarity):
# Kernel messages are first, stored in the kernel # file, critical messages and higher ones also go # to another host and to the console # kern.* /var/log/kernel kern.crit @finlandia kern.crit /dev/console kern.info;kern.!err /var/log/kernel-info
The text further suggests physically printing security information on an old line printer to prevent hackers from erasing their tracks. While that is good advice, it is not so easy to find an inexpensive true ``line printer'' these days. For most environments, a new dot matrix printer would likely suffice (but note that these are becoming more scarce also.) Continuous sheet printer paper is still available, but is now more expensive than stock lettersize or A4 paper.
USAH on page 660 mention secure terminals. However, physical implementations are now archaic; even finding a serial connector on many machines is becoming less common, though not yet rare.
[If you do find such a setup with serial ports, generally the serial ports are connected to a ``terminal server'' or to a true switch (or even cascade of switches) in order to provide more convenient access to many servers.]
In a word, ``don't''.
These were bad ideas from the time that they originated, and they are completely unnecessary these days. Indeed, finding software that even will use these is becoming very rare.
Instead, you should use ssh keys. This is very easy to do; in fact, you can do as simply as
[langley@host1 Slides]$ ssh-keygen -t dsa Generating public/private dsa key pair. Enter file in which to save the key (/home/langley/.ssh/id_dsa): Enter passphrase (empty for no passphrase): [ some passphrase ] Enter same passphrase again: Your identification has been saved in /home/langley/.ssh/id_dsa. Your public key has been saved in /home/langley/.ssh/id_dsa.pub. The key fingerprint is: bb:5b:f6:c4:ed:1b:32:74:90:12:30:ab:60:fd:4b:66 langley@host1 [langley@host1 Slides]$ scp ~/.ssh/id_dsa.pub host2:.ssh/authorized_keys2 The authenticity of host 'host1 (128.186.120.121)' can't be established. RSA key fingerprint is d1:99:0b:9c:b1:ce:87:7d:b7:8e:9a:b5:f1:aa:bc:b9. Are you sure you want to continue connecting (yes/no)? yes Warning: Permanently added 'host2,128.186.120.121' (RSA) to the list of known hosts. langley@host2's password: [ use your password for host2 this time, not your new passphrase ] id_dsa.pub 100% 615 7.1MB/s 00:00 [langley@host1 Slides]$ ssh host2 Enter passphrase for key '/home/langley/.ssh/id_dsa': [ some passphrase ]
So make sure that telnetd, rshd, and rlogind are disabled.
NFS was not designed for security, and you must limit its deployment within strong firewalls. As the text suggests, you should use access lists with fully qualified domain names (or ip numbers.) Squashing ids such as root is a very good idea when you export. When mount, setting ``nosuid'' is also a good idea.
An even better idea is to move users to ``sshfs''; not only does that reduce the amount of configuration that you have to do (simply enable ``sshd''), it makes it very convenient for users to maintain their own mounts.
If you really want to do ``best practices'' with mail security, running postfix or qmail, which were designed with security in mind is probably ``best practice.'' But if you do run sendmail then you should stay abreast of any patches that come out for sendmail. If you are on a mainline Linux distribution such as RedHat/CentOS, just doing a regular yum should be sufficient.
Two of the most useful programs for network security are nmap and zmap. Each can scan machines to see what services might be available. It can search large areas of a network for live machines. As the nmap man page says:
DESCRIPTION
Nmap is designed to allow system administrators and curious individuals
to scan large networks to determine which hosts are up and what ser-
vices they are offering. nmap supports a large number of scanning
techniques such as: UDP, TCP connect(), TCP SYN (half open), ftp proxy
(bounce attack), Reverse-ident, ICMP (ping sweep), FIN, ACK sweep, Xmas
Tree, SYN sweep, IP Protocol, and Null scan. See the Scan Types sec-
tion for more details. nmap also offers a number of advanced features
such as remote OS detection via TCP/IP fingerprinting, stealth scan-
ning, dynamic delay and retransmission calculations, parallel scanning,
detection of down hosts via parallel pings, decoy scanning, port fil-
tering detection, direct (non-portmapper) RPC scanning, fragmentation
scanning, and flexible target and port specification.
Significant effort has been put into decent nmap performance for non-
root users. Unfortunately, many critical kernel interfaces (such as
raw sockets) require root privileges. nmap should be run as root when-
ever possible (not setuid root, of course).
The result of running nmap is usually a list of interesting ports on
the machine(s) being scanned (if any). Nmap always gives the port’s
"well known" service name (if any), number, state, and protocol. The
state is either ’open’, ´filtered´, or ´unfiltered´. Open means that
the target machine will accept() connections on that port. Filtered
means that a firewall, filter, or other network obstacle is covering
the port and preventing nmap from determining whether the port is open.
Unfiltered means that the port is known by nmap to be closed and no
firewall/filter seems to be interfering with nmap’s attempts to deter-
mine this. Unfiltered ports are the common case and are only shown
when most of the scanned ports are in the filtered state.
Depending on options used, nmap may also report the following charac-
teristics of the remote host: OS in use, TCP sequencability, usernames
running the programs which have bound to each port, the DNS name,
whether the host is a smurf address, and a few others.
nmap in its most basic mode is usually quite fast (but zmap is much, much faster):
# nmap diablo Starting nmap V. 3.00 ( www.insecure.org/nmap/ ) Interesting ports on diablo.cs.fsu.edu (128.186.120.2): (The 1595 ports scanned but not shown below are in state: closed) Port State Service 22/tcp open ssh 111/tcp open sunrpc 139/tcp open netbios-ssn 445/tcp open microsoft-ds 515/tcp open printer 4000/tcp open remoteanything Nmap run completed -- 1 IP address (1 host up) scanned in 7 seconds
Here's what mail.cs.fsu.edu looks like:
# nmap mail.cs.fsu.edu Starting nmap V. 3.00 ( www.insecure.org/nmap/ ) Interesting ports on mail.cs.fsu.edu (128.186.120.4): (The 1590 ports scanned but not shown below are in state: closed) Port State Service 22/tcp open ssh 25/tcp open smtp 80/tcp open http 111/tcp open sunrpc 143/tcp open imap2 652/tcp open unknown 847/tcp open unknown 858/tcp open unknown 993/tcp open imaps 995/tcp open pop3s 2049/tcp open nfs Nmap run completed -- 1 IP address (1 host up) scanned in 4 seconds
Nessus (http://www.nessus.org) has matured a lot since USAH wrote their somewhat dismissive blurb on page 665. In fact, it has gone from being open source to closed source just recently. (There is an open source fork project at http://www.openvas.org).
It is a full featured vulnerability scanner that is very popular for penetration testing. Unlike the older Satan and SAINT programs (though I see that SAINT appears to still be alive as a commercial product), Nessus uses its own X client (based on GTK) rather than a web interface.
Metasploit (http://www.metasploit.com/) has been around for many years. It's a framework for exploit management; it lets the community publish exploits in a uniform format for penetration testing.
In the area of general system auditing, there is a program lynis.
Similar in concept, though, are the more specific programs to check for rootkits, such as "chkrootkit" and "rkhunter".
In today's normal environment, you should try to run firewalls such as iptables on all of your servers.
As USAH puts it on page 677, computer server firewalls are, like the firewall in your car, are not a primary means of defense and should not lull a system administrator into a false sense of security. You should have also firewalls that protect your whole site, and in large sites, it is likely that firewalls should be established between production, q/a, and development. You need to continue using other tools that we have discussed, such as nessus and tripwire to look for vulnerabilities, both from the outside of your network (aka ``outside penetration testing'') and from inside your network's firewalls.
As we have discussed before, configuring iptables is not hard; an example set of rules would look something like:
-A INPUT -j RH-Firewall-1-INPUT -A FORWARD -j RH-Firewall-1-INPUT -A RH-Firewall-1-INPUT -i lo -j ACCEPT -A RH-Firewall-1-INPUT -p icmp --icmp-type any -j ACCEPT -A RH-Firewall-1-INPUT -p 50 -j ACCEPT -A RH-Firewall-1-INPUT -p 51 -j ACCEPT -A RH-Firewall-1-INPUT -p tcp --destination-port 25 -j ACCEPT -A RH-Firewall-1-INPUT -m state --state ESTABLISHED,RELATED -j ACCEPT -A RH-Firewall-1-INPUT -j REJECT --reject-with icmp-host-prohibited
Meeting at the juncture of firewalls and IDS are the idea of "intelligent" firewall management.
Here, instead of simply blocking connectivity via a firewall, we allow connectivity until it is abused. Once the IDS side detects abuse, the firewall side is activated, closing the undesired abusive connection.
While there are a number of products that do this at both the network level and the system level, one of my current favorites is "fail2ban". Like the also popular "DenyHost" program, its IDS side is based on scanning logs. Unlike DenyHost, though, it is much more flexible and is designed to handle numerous services.
The idea is that the system administrator has a number of defined "filters" that detect undesired behavior, and a set of actions to apply when problems are detected.
Here is a filter that I created to detect when Dovecot 1.07 is having a problem with a brute-force attack:
[Definition] failregex = pop3-login: Aborted login: .*rip=::ffff:, lip=::ffff:.* imap-login: Aborted login: .*rip=::ffff: , lip=::ffff:.* ignoreregex =
In order to use the filter, you have to include it in your "jail.local" file:
#
# Local stuff
#
[dovecot-iptables]
enabled = true
filter = dovecot
action = iptables-multiport[name=dovecot, port="pop3s,imaps", protocol=tcp]
sendmail[name=dovecot, dest=root]
logpath = /var/log/dovecot.log
maxretry = 20
findtime = 600
bantime = 1200
So what is really going on? How does fail2ban do its work?
It actively manipulates iptables. When it is first started, it creates a new ipchain:
# iptables -L -n Chain INPUT (policy ACCEPT) target prot opt source destination fail2ban-dovecot tcp -- 0.0.0.0/0 0.0.0.0/0 multiport dports 995,993 ACCEPT all -- 0.0.0.0/0 0.0.0.0/0 REJECT all -- 0.0.0.0/0 127.0.0.0/8 reject-with icmp-port-unreachable ACCEPT all -- 0.0.0.0/0 0.0.0.0/0 state RELATED,ESTABLISHED ACCEPT tcp -- 0.0.0.0/0 0.0.0.0/0 tcp dpt:993 ACCEPT tcp -- 0.0.0.0/0 0.0.0.0/0 tcp dpt:995 REJECT all -- 0.0.0.0/0 0.0.0.0/0 reject-with icmp-port-unreachable
Of course, when the ipchain is first created, it has no members:
target prot opt source destination RETURN all -- 0.0.0.0/0 0.0.0.0/0
However, whenever lines matching the "failregex" is detected in /var/log/dovecot.log, entries are added to that ipchain, dropping all packets from that source:
Chain fail2ban-dovecot (1 references) target prot opt source destination DROP all -- 67.192.117.129 0.0.0.0/0 DROP all -- 67.192.117.139 0.0.0.0/0 RETURN all -- 0.0.0.0/0 0.0.0.0/0
You can find more about current security issues via
Some of these give more information, some give less. Here's a typical CERT report on a sendmail problem:
Sendmail Race Condition Vulnerability
Original release date: March 22, 2006
Last revised: --
Source: US-CERT
Systems Affected
Sendmail versions prior to 8.13.6.
Overview
A race condition in Sendmail may allow a remote attacker to
execute arbitrary code.
I. Description
Sendmail contains a race condition caused by the improper handling of
asynchronous signals. In particular, by forcing the SMTP server to
have an I/O timeout at exactly the correct instant, an attacker may
be able to execute arbitrary code with the privileges of the
Sendmail process.
Details, including statements from affected vendors are available
in the following Vulnerability Note:
VU#834865 - Sendmail contains a race condition
A race condition in Sendmail may allow a remote attacker to
execute arbitrary code. (CVE-2006-0058)
Please refer to the Sendmail MTA Security Vulnerability Advisory
and the Sendmail version 8.13.6 release page for more information.
II. Impact
A remote, unauthenticated attacker could execute arbitrary code
with the privileges of the Sendmail process. If Sendmail is
running as root, the attacker could take complete control of an
affected system.
III. Solution
Upgrade Sendmail
Sendmail version 8.13.6 has been released to correct this issue.
In addition to VU#834865, Sendmail 8.13.6 addresses other
security issues and potential weaknesses in the Sendmail code.
Patches to correct this issue in Sendmail versions 8.12.11 and
8.13.5 are also available.
Appendix A. References
* US-CERT Vulnerability Note VU#834865 -
* Sendmail version 8.13.6 -
* Sendmail MTA Security Vulnerability Advisory -
* Sendmail version 8.12.11 Patch -
* Sendmail version 8.13.5 Patch -
* CVE-2006-0058 -
Human factors to consider
Security and obscurity:
Best practices for rolling out new security patches:
Intrusion detection:
USAH on pp. 680-681 gives these steps for handling an attack:
Recovery: Rebuilding your machines?