I highly doubt that 95% are using it for despicable purposes, so I have no ethical concerns with running a Tor relay at home, using Tor regularly, and financially supporting the Tor project.
Why? Well, the most obvious "usable for despicable purposes" feature is .onion services. Today, .onion service traffic makes up about 2% of Tor traffic[1] and the largest .onion service by far is Facebook[2] (served at https://facebookcorewwwi.onion/). So even if we assume that all traffic other than to Facebook is for nefarious purposes (which is not correct given that we know there are other very popular and legal services operating as .onion services), you're looking at much less than 2% of all Tor traffic at best being nefarious. I doubt that the same statistic for clearnet traffic is significantly better.
Now, you might argue that most of the clearnet browsing over Tor is nefarious and that focusing on Onion services is a distraction. But that doesn't really match up with reality either. CloudFlare has claimed in the past that 94% of (clearnet) Tor traffic they see is malicious (meaning "DDoS or spam" not "illegal or immoral content")[3] but they provided no justification for this figure, and at the same time Akamai published a study[4] which found that there is no statistically significant difference when it comes to e-commerce behaviour when comparing Tor users and the regular internet. The Tor Project postulated[5] that this claim by CloudFlare was based on them marking exit nodes as being malicious rather than individual connections from exit nodes and CloudFlare hasn't really responded to that in the past 4 years.
My point is that the 95% figure you posited doesn't pass the sniff test. That means that 1.9 million of the daily users of Tor are all using it for nefarious purposes. If you just count a handful of countries with well-known internet censorship (Brazil, Venezuela, Iran, Egypt, and China) you already have passed the 5% mark of daily Tor users.
[1]: If you compare the onion service and total daily Tor traffic (https://metrics.torproject.org/), you find that it's about 4 Gbit/s out of a total 200 Gbit/s -- so around 2% at time of writing.
Why? Well, the most obvious "usable for despicable purposes" feature is .onion services. Today, .onion service traffic makes up about 2% of Tor traffic[1] and the largest .onion service by far is Facebook[2] (served at https://facebookcorewwwi.onion/). So even if we assume that all traffic other than to Facebook is for nefarious purposes (which is not correct given that we know there are other very popular and legal services operating as .onion services), you're looking at much less than 2% of all Tor traffic at best being nefarious. I doubt that the same statistic for clearnet traffic is significantly better.
Now, you might argue that most of the clearnet browsing over Tor is nefarious and that focusing on Onion services is a distraction. But that doesn't really match up with reality either. CloudFlare has claimed in the past that 94% of (clearnet) Tor traffic they see is malicious (meaning "DDoS or spam" not "illegal or immoral content")[3] but they provided no justification for this figure, and at the same time Akamai published a study[4] which found that there is no statistically significant difference when it comes to e-commerce behaviour when comparing Tor users and the regular internet. The Tor Project postulated[5] that this claim by CloudFlare was based on them marking exit nodes as being malicious rather than individual connections from exit nodes and CloudFlare hasn't really responded to that in the past 4 years.
My point is that the 95% figure you posited doesn't pass the sniff test. That means that 1.9 million of the daily users of Tor are all using it for nefarious purposes. If you just count a handful of countries with well-known internet censorship (Brazil, Venezuela, Iran, Egypt, and China) you already have passed the 5% mark of daily Tor users.
[1]: If you compare the onion service and total daily Tor traffic (https://metrics.torproject.org/), you find that it's about 4 Gbit/s out of a total 200 Gbit/s -- so around 2% at time of writing.
[2]: https://www.facebook.com/notes/facebook-over-tor/1-million-p...
[3]: https://blog.cloudflare.com/the-trouble-with-tor/
[4]: https://www.stateoftheinternet.com/downloads/pdfs/state-of-t...
[5]: https://blog.torproject.org/trouble-cloudflare