Denial of Wallet Attacks & The Case for Anonymity
While looking into potentially migrating away from running my own DNS and mail server I looked into which providers I was already implicitly depending on based on services I trust. In doing this, I began to settle on Route 53 for DNS hosting. However, Route53 bills per query. A very small amount per query, but per query nevertheless. This got me asking: what if somebody were to deliberately query me as many times as possible?
Looking at a sample PCAP I took while my server was under a DNS ANY query attack (open Statistics > DNS in Wireshark and use the display filter “dns.qry.type ANY and dns.flags 0x0010”), the DNS statistics show 80,355 ANY queries at a rate of 1.2814 per millisecond from between 13:03:46 to 13:04:49. Assuming a constant rate for 1 week this would be 826,656,768 ($330.66 USD) and over 1 month that would be 3,592,020,480 queries ($918.40 USD) (math here if you're interested). Route 53 do have mitigations in place that kick in for free, but to some degree you are reliant on them to let in just the right kind of technically-compliant traffic and blocking what they think to be malicious based on pattern recognition, meaning some will get through. This is great if you’re running a business that must stay up and is likely the one of the reasons Route 53 is used by many of the services I depend on, but for some of us, just going offline is better than getting a huge bill.
In my case, this attack did take down my firewall for some time despite being a pretty modest amount of bandwidth because it caused the firewall to run out of sessions. The exhaustion issue is trivially fixed on my current setup by increasing the amount of sessions and RAM (and later applying some IDR rules and DNS configuration tweaks ) but shows how just one well-resourced attacker with the ability to spoof IPs can cause thousands of dollars in damage using just one kind of attack. Say you block this attack, what if they start flooding you with UDP bandwidth, and more importantly, what about those close to you?
Say I had the most bulletproof setup possible for myself. Low cost, high uptime and high-resilience to DDoSs (which is something you can probably achieve using Cloudflare for many use cases). The next move (if it wasn’t the first) would be to go after your employer. You might have a rock-solid setup and can afford to go down, but what about them? Would they stick by their guns and keep you around if getting rid of you is cheaper than fighting off the attack? It’s not just about technical mitigations at that point. Once the target has been identified, they have already lost given a determined enough threat. This isn’t much of a hypothetical either; gaming communities used to get into fights and DDoS each other to destroy their player bases and try and siphon some of those players off to themselves. DDoSs are the oldest feuds on the Internet.
Ultimately, the best mitigation for DDoSs and other semi-anonymous attacks is to not become a target in the first place. If you think what you are about to do has a reasonable chance of setting off the wrong person (somebody petty enough to act on an Internet grudge that’s resourceful enough to do it) then wear the veil of anonymity, at least where it makes sense. Posting/commenting about a contentious issue on social media? Consider a separate account and Tor. Petty gaming communities? Maybe a VPN on your router and a separate account (though this is an opsec nightmare in practice). The Internet can be one of the most privacy-invasive and insecure places on each, but also the least.