rdiscd has been designed and implemented with security in mind.  Safe,
high-level C++ constructs are used whenever possible to reduce the chance
of programmer error.  Low-level, potentially-unsafe code is kept to a
minimum and written very carefully when it's necessary.  In addition,
I have audited the code of libndp, the NDP library used by rdiscd. I
found one vulnerability, CVE-2014-3554, which is fixed as of libndp 1.4.

By necessity, rdiscd requires privilege, as it configures the system's
networking and uses raw sockets.  To mitigate the potential impact of
a security vulnerability in rdiscd or libndp, rdiscd supports privilege
separation.  The process that handles NDP packets from the network runs
as an unprivileged user in an empty and unwritable root directory, and
passes configuration information to the privileged process over a UNIX
domain socket using a simple and easily-audited protocol.  The privileged
process never talks directly to the network.

To use privilege separation, you must pass the name of an unprivileged
user to the -u option (you should use a dedicated user for rdiscd),
and a root directory to the -r option (you should use an empty and
unwritable directory).  Privilege separation is enabled by default in
the Debian package, with no manual configuration required.

As of 2021, rdiscd is no longer maintained and security vulnerabilities
will not be fixed.  If you find a security vulnerability in rdiscd,
please report it to a full disclosure mailing list so users of rdiscd
are made aware of it.