pretty basic question but i was wondering if there were any larger risks associated with self hosting a tor website, like is there anything in particularly to watch out for
i live in a country without much risks (no us agencies/europol involvement where i am), and i'm already taking all other opsec practices normally
This post was migrated here from /d/OpSec
This comment was posted automatically by a bot. All AutoModerator settings are configured by individual communities. Contact this community's Moderators to have your post approved if you believe this was in error.
This comment was posted automatically by a bot. All AutoModerator settings are configured by individual communities. Contact this community's Moderators to have your post approved if you believe this was in error.
If your website doesn't have proper measures against bad actors, then someone could find their way inside and cause major headaches for you. Pentest your site! Or pay someone else to do it.
Can share my experience with /d/pgpfreaks, which is a non-monetized and arguably, low-risk service (compared to markets or exchanges anyway). Over the course of the first month, I've been under attack half a dozen times. As a clarification to our users, we do not keep logs as mentioned in our ToS but I have a couple of triggers enabled to detect unusual activity. I do not exclude some of the attempts went under the radar, too.
Weaknesses in Software Stack (0-Day)
You may have hardened your Nginx or Linux to the maximum. But if there is a 0-day exploit in the libevent library or in the Tor source code itself, even the best opsec won't help. Attackers can get Remote Code Execution (RCE) directly through your onion route.
Guard Discovery Attack
This is a long-term threat. Tor uses an “Entry Guard” system (the first node you connect to). The problem is, if your server stays on for months or years, there is a possibility that enemies with large resources (at the state or large ISP level) can perform statistical analysis to find out who your Guard is. Once your Guard is discovered, your real identity is just one step away from being exposed.
DDoS at the Tor Layer (Circuit Stressing)
This is different from a regular internet DDoS. Attackers don't need large bandwidth to shut down your website. They can spam circuit creation requests to your website's Introduction Point. Because the cryptographic process of opening a Tor path is CPU-intensive, your server can become overloaded just from handling the “handshakes” from thousands of fake circuits.
Correlation Attack (Traffic Analysis)
If you get enough traffic, an enemy who can monitor the “end” (your home internet) and the “source” (traffic coming out of a specific Tor node) can match the timing patterns of incoming and outgoing data.
The problem: Tor is low-latency, so it doesn't add enough delay to fool timing analysis when the data is really big.
You may have hardened your Nginx or Linux to the maximum. But if there is a 0-day exploit in the libevent library or in the Tor source code itself, even the best opsec won't help. Attackers can get Remote Code Execution (RCE) directly through your onion route.
Guard Discovery Attack
This is a long-term threat. Tor uses an “Entry Guard” system (the first node you connect to). The problem is, if your server stays on for months or years, there is a possibility that enemies with large resources (at the state or large ISP level) can perform statistical analysis to find out who your Guard is. Once your Guard is discovered, your real identity is just one step away from being exposed.
DDoS at the Tor Layer (Circuit Stressing)
This is different from a regular internet DDoS. Attackers don't need large bandwidth to shut down your website. They can spam circuit creation requests to your website's Introduction Point. Because the cryptographic process of opening a Tor path is CPU-intensive, your server can become overloaded just from handling the “handshakes” from thousands of fake circuits.
Correlation Attack (Traffic Analysis)
If you get enough traffic, an enemy who can monitor the “end” (your home internet) and the “source” (traffic coming out of a specific Tor node) can match the timing patterns of incoming and outgoing data.
The problem: Tor is low-latency, so it doesn't add enough delay to fool timing analysis when the data is really big.