scanlogd - detects and logs TCP port scans

scanlogd

scanlogd detects port scans and writes one line per scan via the syslog(3) mechanism. If a source address sends multiple packets to different ports in a short time, the event will be logged. The format of the messages is:

saddr[:sport] to daddr [and others,] ports port[, port...], ..., flags[, TOS TOS][, TTL TTL] @HH:MM:SS

The fields in square brackets are optional; sport, TOS, and TTL will only be displayed if they were constant during the scan.

The flags field represents TCP control bits seen in packets coming to the system from the address of the scan. It is a combination of eight characters, with each corresponding to one of the six defined and two reserved TCP control bits (see RFC 793). Control bits that were always set are encoded with an uppercase letter, and a lowercase letter is used if the bit was always clear. A question mark is used to indicate bits that changed from packet to packet.

In order to do its job, scanlogd needs a way to obtain raw IP packets that either come to the system scanlogd is running on, or travel across a network segment that is directly connected to the system. Current versions of scanlogd can be built with support for one of several packet capture interfaces.

scanlogd is aware of the raw socket interface on Linux, libnids, and libpcap.

The use of libpcap alone is discouraged. If you're on a system other than Linux and/or want to monitor the traffic of an entire network at once, you should be using libnids in order to handle fragmented IP packets.

At least 7 different privileged or 21 non-privileged ports, or a weighted combination of those, have to be accessed with no longer than 3 seconds between the accesses to be treated as a scan. If more than 5 scans are detected within 20 seconds, that event will be logged and logging will be stopped temporarily.

Logging is done with a facility of daemon and a priority level alert.

scanlogd should be started as root since it needs access to a packet capture interface. By default, it chroots to /var/empty and switches to running as user scanlogd after the packet capture interface is initialized.

If the daemon couldn't start up successfully, it will exit with a status of 1.

You're expected to create a dummy user for scanlogd to run as. Make sure you allocate unique UID and GID to the user.

In most cases, scanlogd should be started from a rc.d script on system startup.

In /etc/syslog.conf you may use something like:

daemon.alert /var/log/alert

As the name indicates, scanlogd only logs port scans. It does not prevent them. You will only receive summarized information in the system's log.

Obviously, the source address of port scans can be spoofed. Don't take any action against the source of attacks unless other evidence is available. Sometimes IP addresses are shared between many people; this is the case for ISP shell servers, dynamic dialup pools, and corporate networks behind NAT (masquerading).

Due to the nature of port scans, both false positives (detecting a scan when there isn't one) and false negatives (not detecting a scan when there's one) are possible. In particular, false positives occur when many small files are transferred rapidly with passive mode FTP.

Solar Designer <solar at openwall.com>

Steffen Dettmer <steffen at dett.de> wrote the initial version of this manual page.

syslog(3), syslog.conf(5), libnids(3), pcap(3)

scanlogd home page: http://www.openwall.com/scanlogd/
Phrack Magazine, issue 53, article 13