Sentry

NAME
sentry - safe and effective protection against bruteforce attacks

DESCRIPTION
Sentry detects and prevents bruteforce attacks against sshd using minimal system resources.

SAFE
To prevent inadvertant lockouts, Sentry manages a whitelist of IPs that have connected more than 3 times and succeeded at least once. Never again will that forgetful colleague behind the office NAT router get us locked out of our system. Nor the admin whose script just failed to login 12 times in 2 seconds.

Sentry includes support for adding IPs to a firewall. Support for IPFW, PF, ipchains is included. Firewall support is disabled by default. This is because firewall rules may terminate existing session(s) to the host (attn IPFW users). Get your IPs whitelisted (connect 3x or use --whitelist) before enabling the firewall option.

SIMPLE
Sentry has an extremely simple database for tracking IPs. This makes it very easy for administrators to view and manipulate the database using shell commands and scripts. See the EXAMPLES section.

Sentry is written in perl, which is installed everywhere you find sshd. It has no dependencies. Installation and deployment is extremely simple.

FLEXIBLE
Sentry supports blocking connection attempts using tcpwrappers and several popular firewalls. It is easy to extend sentry to support additional blocking lists.

Sentry was written to protect the SSH daemon but anticipates use with other daemons. SMTP support is planned. As this was written, the primary attack platform in use is bot nets comprised of exploited PCs on high-speed internet connections. These bots are used for carrying out SSH attacks as well as spam delivery. Blocking bots prevents multiple attack vectors.

The programming style of sentry makes it easy to insert code for additonal functionality.

EFFICIENT
The primary goal of Sentry is to minimize the resources an attacker can steal, while consuming minimal resources itself. Most bruteforce blocking apps (denyhosts, fail2ban, sshdfilter) expect to run as a daemon, tailing a log file. That requires a language interpreter to always be running, consuming at least 10MB of RAM. A single hardware node with dozens of virtual servers will lose hundreds of megs to daemon protection.

Sentry uses resources only when connections are made. The worse case scenario is the first connection made by an IP, since it will invoke a perl interpreter. For most connections, Sentry will append a timestamp to a file, stat for the presense of another file and exit.

Once an IP is blacklisted for abuse, whether by tcpd or a firewall, the resources it can consume are practically zero.

Sentry is not particularly efficient for reporting. The "one file per IP" is superbly minimal for logging and blacklisting, but nearly any database would perform better for reporting. Expect to wait a few seconds for sentry --report.

REQUIRED ARGUMENTS

 * ip
 * An IPv4 address. The IP should come from a reliable source that is difficult to spoof. Tcpwrappers is an excellent source. UDP connections are a poor source as they are easily spoofed. The log files of TCP daemons can be good source if they are parsed carefully to avoid log injection attacks.

All actions except report and help require an IP address. The IP address can be manually specified by an administrator, or preferably passed in by a TCP server such as tcpd (tcpwrappers), inetd, or tcpserver (daemontools).

ACTIONS

 * blacklist
 * deny all future connections


 * whitelist
 * whitelist all future connections, remove the IP from the blacklists, and make it immune to future connection tests.


 * delist
 * remove an IP from the white and blacklists. This is useful for testing that sentry is working as expected.


 * connect
 * register a connection by an IP. The connect method will log the attempt and the time. See CONNECT.


 * update
 * Check the most recent version of sentry against the installed version and update if a newer version is available.

SHELL COMMANDS
View the total number of connections:

the number of unique IPs that have connected:

the timestamps for every connection 10.0.1.193 made:

check if 10.0.1.193 is whitelisted

NAUGHTY
Sentry has flexible rules for what constitutes a naughty connection. For SSH, attempts to log in as an invalid user are considered naughty. For SMTP, the sending of a virus, or an email with a high spam score could be considered naughty. See the configuration section in the script related settings.

CONNECT
When new connections arrive, the connect method will log the attempt and the time. If the IP is white or blacklisted, it will exit immediately.

Next, sentry checks to see if it has seen the IP more than 3 times. If so, check the logs for successful, failed, and naughty attempts from that IP. If there are any successful logins, whitelist the IP and exit.

If there are no successful logins and there are naughty ones, blacklist the IP. If there are no successful and no naughty attempts but more than 10 connection attempts, blacklist the IP. See also NAUGHTY.

CONFIGURATION AND ENVIRONMENT
There is a very brief configuration section at the top of the script. Once your IP is whitelisted, update the booleans for your firewall preference and Sentry will update your firewall too.

Sentry does NOT make changes to your firewall configuration. It merely adds IPs to a table/list/chain. It does this dynamically and it is up to the firewall administrator to add a rule that does whatever you'd like with the IPs in the sentry table.

I use the sentry IP table like so with PF:

That blocks all connections from anyone in the sentry table.

DIAGNOSTICS
Sentry can be run with --verbose which will print informational messages as it runs.

DEPENDENCIES
Sentry uses only modules built into perl. Additional modules may be used in the future but Sentry will not depend upon them. In other words, if you extend Sentry with modules are aren't built-ins, also include a fallback method.