Essay:On Internet censorship
In this essay, I intend to talk about Internet censorship and my stance towards some of the ways this is accomplished.
This form of censorship relies on modifying the responses from a DNS server to return predetermined responses for certain domains instead of returning the data received from authoritative nameservers. Commonly used when ISPs receive orders to cut access to torrent index websites, this attack is possible because one of the services ISPs offer to subscribers is a recursive DNS server. Not as common nowadays because ISPs increasingly "outsource" this task to DNS resolvers like Google's 188.8.131.52/2001:4860:4860::8888, though even if they outsource it is still technically possible to tamper with the answers a subscriber receive from these servers because the DNS protocol is by default unencrypted.
Torrent sites aren't the only victim of tampering with the DNS system. In Spain, during the Catalonian independence referendum, domains related to the referendum were subject to censorship through tampering DNS responses (for domains outside of .cat, which they seized straight up).
I'm not a fan of this. DNS infrastructure is central to the Internet, and tampering with it like this is something I don't like and consider an attack on the Internet, especially when it gets into censorship of the type that Spain did during the referendum in Catalonia, that is, censorship of an explicit political nature (though I consider censorship of torrent sites of a political nature as well). I believe recursive resolvers should strictly return the data received from authoritative name servers when querying for a public domain name (excluding TLDs like .local)
This is also a pretty ineffective form of censorship, as all it takes is to defeat it is to change DNS servers and/or setup DoT/DoH.
I believe there should be strict legislation around this to the point that governments are effectively forbidden from forcing ISPs to perform this attack. I also believe these are online intermediaries that should be protected already from liability from content published in domains that they resolve.
There are a series of routers between a end-user's computer and the computer/server it wants to connect to (you can view them using the traceroute command in Linux). These routers are in charge of carrying your packets from your computer to the computer it wants to connect to.
This opens the possibility of censorship through dropping packets intended to go to certain addresses. Used especially when censoring the Tor network. [note 1]
Much like in the DNS case, I consider the routers as intermediaries without liabilities about which IP addresses communicate with which IP address. I also consider making these devices drop packets like this an attack on the Internet on an even more concerning scale that the DNS case. Also like in the DNS case, I believe they should just concern themselves with carrying the packets from one computer to the other without discriminating packets based on what addresses are involved
Private companies operating sites which post user generated content deleting/restricting said content
When you hear right-wingers talking about Internet censorship, they mostly refer to things like Facebook/Twitter banning them and Youtube deleting or putting age restrictions on their videos.
Unlike the above 2 cases, these platforms, however big they are, are not vital internet infrastructure. Like Eugen Rochko, I don't consider these services public utilities either. Concerning content that users post to their websites with the intent of making it public and distributing it, they are free to either host it or delete it and ban the user for all I care.
- ↑ 1.0 1.1 https://ooni.org/post/internet-censorship-catalonia-independence-referendum/
- ↑ https://www.quad9.net/news/blog/quad9-files-official-objection-opposing-sony-music-s-german-court-ruling/
- ↑ https://explorer.ooni.org/measurement/20220818T235325Z_tor_RU_41727_n1_aTJGoxA7Vzf0SFqS
- ↑ https://ooni.org/post/2021-russia-blocks-tor/
- ↑ In this test, the dropping packets is seen as "generic_timeout_error". There's no way for the software performing the test to know whether this is because the tor node is offline or busy, or because of deliberate censorship, thus it is marked as an anomaly instead of confirmed, however OONI already verified deliberate censorship of the Tor network since December 2021 in a separate blog post. One of the connections fails with "network_unreachable" because the device doesn't have access to the IPv6 network, most likely because their ISP doesn't yet support it (we are in 2022 god dammit)