The major attack of concern described in the paper is the transparent early terminated encryption attack, and root trust signing that fall under effectively the

The major attack of concern described in the paper is the transparent early terminated encryption attack, and root trust signing that fall under effectively the same centralized hands at the AS level. Where an AS level entity MITMs all outbound connections from a region in automated fashion for collection, before that traffic ever makes it to TOR or its destination. It works for TOR, TLS, pretty much any protocol out there where key exchange or trust occurs; so long as the protocol is known and has distinct classifiable characteristics allowing computation to automatically do this. There have been instances where public certs issued by a CA with the same domain names, but are issued from a root CA that is other than the legitimate site's root CA which are used for attacks. CT logs don't stop this either. There is a lot of ephemeral content, and private information that can be both collected, and injected on a targeted basis if one has access to such junctions which the industry (Telecom) has proven time and again that they can't secure following basic practice; largely because mandates to backwards compatibility at the regulatory level. Social credit, where invisible factors people don't control force those same people into poverty through targeted denial of service (communications for job hunting/social contacts), zersetzung, etc; that all would be a breeze to set up without any external indicator, or remedy using that attack. What the target sees vs what everyone else sees would be quite different, and of course there would be people that gaslight and torture on top of it all (as a natural psychological defense mechanism of denial). Compromised communications under such type of attacks are madness inducing.