As anyone who's been in the DDoS attack trenches knows, large multi-gigabit attacks have become more prevalent over the last few years. For many organizations, it's become economically unfeasible to provision enough bandwidth to combat this threat.
How are attackers themselves sourcing so much bandwidth? It's actually easier than you might think. While botnets comprised of malware-infected computers can be used to launch attacks, you don't actually need thousands of devices. In some cases, attackers are infiltrating hosting company resources (shared hosting, virtual private servers, dedicated hosting, etc.), availing themselves of bandwidth by using hacked, stolen and fraudulent accounts.
Let's say that an attacker manages to get his/her hands on 5 hosting accounts with 5 different hosting companies. It's not unusual for these hosting companies to have 1 Gbps+ of connectivity to the Internet. A lot of hosters don't look at their outbound traffic all that closely or have difficulty policing what their customers do. All an attacker needs to do is install a script on each account and he/she has easy access to gigabits of connectivity.
For hosters, finding the trouble spot can be like looking for a needle in a haystack (especially if thousands of accounts share resources). While the offender might be found eventually and the account shut down, the damage has already been done.
What can hosters do to help prevent this or detect this better?
Restrict outbound traffic from your customers by using ACLs (Access Control Lists). For example, there are few reasons your customers will ever need to make port 80 UDP connections to other hosts on the Internet. Put policies in place to block all outbound traffic except to specific, acceptable, understood destinations or ports. If customers have legitimate reasons to make an outbound connection from your infrastructure, they should be able to notify you and justify it (this will affect a only tiny percentage of your base) so you can make the appropriate arrangements. Some hosters do not even accommodate these requests.
Throttle outbound traffic from your customers. Even for legitimate outbound connections, most likely they don't need to take up 500 Mbps of outbound bandwidth. Simply set a lower limit.
Put alarms in place when outbound traffic utilization spikes. If, for example, all of a sudden the amount of data leaving your network increases by 40%, there's probably an issue somewhere and your tech folks should be investigating.
Restricting and monitoring your outbound traffic will probably save you money on bandwidth costs and decrease the amount of abuse reports. Best of all, attackers will realize they're not getting what they want out of your platform. The less you have to worry about, the better, right?
By Miguel Ramos, Sr. Product Manager, Neustar Enterprise Services
|Cybersquatting||Policy & Regulation|
|DNS Security||Registry Services|
|IP Addressing||White Space|
Afilias - Mobile & Web Services
Minds + Machines