What’s stopping government agencies and intelligence groups (such as the 14 eyes) from running a ton of nym nodes and using the info from those to de-anonymize users?
I know that the Tor project has had similar saftey concerns so I was wondering if nym VPN has taken any, steps to root out suspicious node operators or avoid using too many nym nodes potentially owned by the same group.
hello asfaik there where 2 papers
which adresses some mass atacks on the nym network:
these 2 papers indicates imho that the nym network has an good robust layer preventing these attacks.
but i think if u have control over an significant percent amount of nodes u are able to rum some more sophisticated attacks
moreover asfaik there were also no real live Independent mass node tests for deanomyzation,mass manipulation available for now.
may in the future i dont know.
note: these is are my personal assumptions.
may the nym team can correct me or add sth.
regards
Privacy or anonymity does cost money and can’t be obtained if you are a passive and consuming user. You need to invest money and run your very own node(s), either entry or exit with your friends.
Secondly, with gag-orders your ISP may have to report that you are
using such technology and then TLAs can look directly at your
activities, in case there is a demand for that. And for that case you
as a user may not know it and how to protect against that.
Thank you for giving such a detailed response! I’ll be sure to look into these.
The network is permissionless so we don’t exactly “vet” operators for trustworthiness or “stop” govt. agencies/intelligence from running nodes. That’d just mean rebuilding the trusted third party the network was actually supposed to remove but there’s more nuance to this. Look below for details.
Deanonymizing a user on the mixnet isn’t a matter of running a lot of nodes. You need the full path.
Traffic in Anonymous Mode takes 5 hops: entry gateway, three mix nodes across three separate layers, exit gateway. Your client picks one node at random from each layer, and it does that per packet, not per session. There is no circuit to compromise. A single stream is scattered across a large number of independent paths, so an adversary who owns your entry and exit sees a fraction of your packets and no way to tell which fraction.
In addition to that: every packet is padded to a uniform size, delays at each hop is drawn from a Poisson distribution, and real traffic is mixed with cover traffic that is indistinguishable from it. That combination is what breaks the timing correlation an entry-and-exit adversary depends on. This is the attack that worked against Tor in the 2014 relay-early case, and it is the attack the mixnet was designed around.
Layer assignment also rotates each epoch, so you can’t quietly park nodes in the positions you want and wait. Per the mixnode deep dive (from the whitepaper), an adversary controlling 50% of nodes observes roughly 12.5% of messages across the three layers, and that number is before cover traffic and mixing delays are accounted for.
Stake makes the attempt expensive. Nodes enter the active set through randomized selection weighted by delegated stake, so “run a ton of nodes” means “bond and attract a lot of NYM on every one of them.” The price of running reputed nodes should also increase as the network sees more demand. The two papers @moejoe linked are the right ones to be reading. There has not been an independent large-scale adversarial test against the live network.
Tor has no agreement with its relay operators. Anyone can run one, and nothing they sign constrains what they do with what passes through.
Nym node operators accept the Node Operators and Validators Terms before they can bond. Section IV is explicit: an operator “will not collect, monitor, record, log, store, retain, or pass on to any third party Nym Node information or any information or data relating to the activities of end users.” Section VI separately prohibits manipulating network responses in ways that compromise the integrity of the mixnet.
That is a signed, enforceable agreement. An agency running Nym nodes to harvest user metadata is knowingly in breach of it. That gives defense counsel a documented basis to challenge how such evidence was obtained. This is something that Tor doesn’t have. We went into this in more detail on a recent community call with former Silk Road 2.0 operator Blake Benthall: https://www.youtube.com/watch?v=8EW2AXNibhc
Like @Ch1ffr3punk mentioned: the strongest answer to “who runs the nodes” is “you do.” 
The network is open, and anyone can bond and run one. You are welcome to use your own nodes. We also work against concentration directly through the Delegations Program. The treasury delegates up to 50% of stake saturation, capped at 125k NYM, to operators who meet the criteria, and it targets locations where the network needs coverage most. As an operator builds independent stake, the treasury delegation steps back down and eventually withdraws. The purpose is to keep the operator base geographically and jurisdictionally diverse, and to make sure a new independent operator can reach the active set without arriving already rich.
Note that none of this rests on trusting Nym, or on trusting any individual operator. That’s the point of having things decentralised and open-source.
Happy to go deeper on any of these. Would also welcome your feedback or any other questions!