Linux Certif

Toute la documentation sur la certification Linux LPI

sshguard

BSD mandoc

NAME

sshguard - monitors daemon activity

SYNOPSIS

[-b thr:filename ] [-a abuse_tresh ] [-p pardon_min_interval ] [-s stale_tresh ] [-w addr/host/block/file ] [-f srv:pidfile ]

DESCRIPTION

monitors logging activity and reacts to attacks by blocking their source addresses.

has born for protecting SSH servers from the today's widespread brute force attacks, and evolved to an extensible log supervisor for blocking attacks to applications in real-time.

is given log messages in its standard input. By means of a parser, it decides whether an entry is normal activity or attack; in the latter case, it remarks the author's address. After a number of attacks, the address is distinguished as an abuser and blocked with the firewall.

supports the following firewalls:

AIX native firewall

for IBM AIX operating systems

netfilter/iptables

for Linux-based operating systems

Packet Filter (PF)

for BSD operating systems (Open, Free, Net, DragonFly-BSD)

IPFirewall (IPFW)

for FreeBSD and Mac OS X

IP Filter (IPFILTER)

for FreeBSD, NetBSD and Solaris

tcpd's hosts_access (/etc/hosts.allow)

portable across UNIX

null

portable do-nothing backend for applying detection but not prevention

USAGE

is given log entries to scan in its standard input. It can interpret logs with the following formats:

• syslog
• syslog-ng
• metalog
• multilog
• raw messages

can interface with the following blocking systems to block attackers:

• IBM AIX firewall
• PF
• netfilter/iptables
• IPFW
• IP Filter
• /etc/hosts.allow
• null firewall

Depending on the way chosen for giving logs to sshguard, and depending on the chosen blocking system, some setup may be needed. Instructions are documented at http://www.sshguard.net/doc/setup/setup.html .

does not make use of any configuration file. Instead, a combination of optional arguments can be passed to its process on the command line, for modifying its default behaviour:

-b [num:]filename

enable blacklisting: blacklist after num (or 3) blocked abuses, and hold the permanent blacklist in filename See TOUCHINESS & BLACKLISTING below.

-v

print summary information on sshguard and exit.

-a num

block an address after num attack attempts have been detected. (Default: 4)

-p secs

release a blocked address not sooner than secs seconds after being blocked. will release the address between X and 3/2 * X seconds. (Default: 7*60)

-s secs

forget about an address after secs seconds. If host A issues one attack every this many seconds, it will never be blocked. (Default: 20*60)

-w addr/host/block/file

see the WHITELISTING section.

-f servicecode:pidfile

see the LOG MESSAGE AUTHENTICATION section.

When is signalled with SIGTSTP, it suspends activity. When is signalled with SIGCONT, it resumes monitoring. During suspension, log entries are discarded without being analyzed.

When senses the SSHGUARD_DEBUG environment variable, it enables debugging mode: logging is directed to standard error instead of syslog, and includes comprehensive details of the activity and parsing process. Debugging mode can help investigating patterns: once enabled, a pattern can be directly pasted into the tool from the console, and the behavior is immediately and minutely shown beneath.

WHITELISTING

supports address whitelisting. Whitelisted addresses are not blocked even if they appear to generate attacks. This is useful for protecting lame LAN users (or external friendly users) from being incidentally blocked.

Whitelist addresses are controlled through the -w command-line option. This option can add explicit addresses, host names and address blocks:

addresses

specify the numeric IPv4 or IPv6 address directly, like:

-w 192.168.1.10

or in multiple occurrences:

-w 192.168.1.10 -w 2001:0db8:85a3:0000:0000:8a2e:0370:7334

host names

specify the host name directly, like:

-w friendhost.enterprise.com

or in multiple occurrences:

-w friendhost.enterprise.com -w friend2.enterprise.com

All IPv4 addresses that the host resolves to are whitelisted. Hosts are resolved to addresses once, when sshguard starts up.

address blocks

specify the address block in the usual CIDR notation:

-w 192.168.0.0/24

or in multiple occurrences:

-w 192.168.0.0/24 -w 1.2.3.128/26

file

When longer lists are needed for whitelisting, they can be wrapped into a plain text file, one address/hostname/block per line, with the same syntax given above.

can take whitelists from files when the -w option argument begins with a `.' (dot) or `/' (slash).

This is a sample whitelist file (say /etc/friends):

 # comment line (a '#' as very first character)
 #   a single IPv4 and IPv6 address
 1.2.3.4
 2001:0db8:85a3:08d3:1319:8a2e:0370:7344
 #   address blocks in CIDR notation
 127.0.0.0/8
 10.11.128.0/17
 192.168.0.0/24
 #   hostnames
 rome-fw.enterprise.com
 hosts.friends.com
 

And this is how is told to make a whitelist up from the /etc/friends file:

sshguard -w /etc/friends

The -w option can be used only once for files. For addresses, host names and address blocks it can be used with any multiplicity, even with mixes of them.

LOG MESSAGE AUTHENTICATION

Syslog and syslog-ng typically insert a PID of the generating process in every log line. This can be checked for authenticating the source of the message and avoid false attacks to be detected because malicious local users inject crafted log lines. This way can be safely used even on hosts where this assumption does not hold.

Log message authentication is only needed when is fed log messages from syslog or from syslog-ng. When a process logs directly to a raw file and sshguard is configured for polling logs directly from it, you only need to adjust the log file permissions so that only root can write on it.

For enabling log message authentication on a given service the -f option is used as follows:

-f 100:/var/run/sshd.pid

which associates the given pidfile to the ssh service (code 100). A list of well-known service codes is available at http://www.sshguard.net/doc/servicecodes.html

The -f option can be used multiple times for associating different services with their pidfile:

sshguard -f 100:/var/run/sshd.pid -f 123:/var/run/mydaemon.pid

Services that are not configured for log message authentication follow a default-allow policy (all of their log messages are accepted by default).

PIDs are checked with the following policy:

1. the logging service is searched in the list of services configured for authentication. If not found, the entry is accepted.
2. the logged PID is compared with the pidfile. If it matches, the entry is accepted
3. the PID is checked for being a direct child of the authoritative process. If it is, the entry is accepted.
4. the entry is ignored.

Low I/O load is committed to the operating system because of an internal caching mechanism. Changes in the pidfile value are handled transparently.

TOUCHINESS & BLACKLISTING

In many cases, attacks against services are performed in bulk in an automated form. For example, the attacker goes trough a dictionary of 150 username/password pairs and sequentially tries to violate the SSH service with any of them, continuing blindly while blocked, and re-appearing once the block expires.

To counteract these cases, by default behaves with touchiness Besides observing abuses from the log activity, it monitors also the overall behavior of attackers. The decision on when and how to block is thus made respective to the entire history of the attacker as well. For example, if address A attacks repeatedly and the base blocking time is 420 seconds, A will be blocked for 420 seconds (7 mins) at the first abuse, 2*420 (14 mins) the second, 2*2*420 (28 mins) the third ... and 2^(n-1)*420 the n-th time.

Touchiness has two major benefits: to legitimate users, it grants forgiving blockings on failed logins; to real attackers, it effectively renders large scale attacks infeasible, because the time to perform it explodes with the number of attempts.

Touchiness can be augmented with blacklisting (-b). With this option, after a number of abuses, the address is added to a list of attackers to be blocked permanently. The list is intended to be loaded at each startup, and maintained/extended with new entries during operation. inserts a new address after it exceeded a threshold of abuses. This threshold is configurable within the -b option argument. Blacklisted addresses are never scheduled for releasing.

The -b command line option enables blacklisting and requires the filename to use for permanent storage of the blacklist. Optionally, a custom blacklist threshold can be prefixed to this path, separated by ':'. For example,

-b 5:/var/db/sshguard/blacklist.db

requires to blacklist addresses after the 5th abuse, and store the blacklist in file /var/db/sshguard/blacklist.db. Although the blacklist file is not meant to be in human-readable format, the strings(1) command can be used to peek in it for listing the blacklisted addresses.

EXTENSIONS

can be easily extended to support both more backends (systems blocking addresses, like firewalls) and to recognize more attack patterns.

Adding backends is extremely easy when the blocking and releasing operations can be controlled with system commands. provides a shell script for generating such extensions in few steps: sshguard_backendgen.sh

Adding more attack patterns needs some expertise with bison, as uses a grammar-based context-free parser for powerfulness. Thus, there is one tracker for user-proposed patterns at http://www.sshguard.net/newattackpatt.php

SEE ALSO

syslog(1), syslog.conf5

website at: http://www.sshguard.net/