A Botnet Accidentally Destroyed I2P | Hacker News Reader

A Botnet Accidentally Destroyed I2P

sambent.com

157 pointsCider998691 comments

This seems to lack the full story, despite the headline.. Krebs' coverage is more in-depth (39 points) https://news.ycombinator.com/item?id=46976825

From the main article, I2P has 55,000 computers, the botnet tried to add 700,000 infected routers to I2P to use it as a backup command-and-control system.

https://news.ycombinator.com/item?id=46976825

This, predictably, broke I2P.

That's an interesting stress test for I2P. They should try to fix that, the protocol should be resilient to such an event. Even if there are 10x more bad nodes than good nodes (assuming they were noncompliant I2P actors based on that thread) the good nodes should still be able to find each other and continue working. To be fair spam will always be a thorny problem in completely decentralized protocols.

No. They should not try to survive such attacks. The best defense to a temporary attack is often to pull the plug. Better than than potentially expose users. When there are 10x as many bad nodes as good, the base protection of any anonymity network is likely compromised. Shut down, survive, and return once the attacker has moved on.

Why would an attacker move on if it can maintain a successful DoS attack forever?

Because botnets are mostly there to make money nowadays. Or owned by state actors.

Either way, it’s opportunity cost.

The mentioned botnet didn't intentionally take down I2P. It's run by bunch of kids who don't know what they're doing.

This is why Tor is centralized, so that they can take action like cutting out malicious nodes if needed. It’s decentralized in the sense that anyone can participate by default.

> so that they can take action like cutting out malicious nodes if needed

How does that work?

While anyone can run a Tor node and register it as available, the tags that Tor relays get assigned and the list of relays is controlled by 9 consensus servers[1] that are run by different members the Tor project (in different countries). They can thus easily block nodes.

[1]: https://consensus-health.torproject.org/

It's 10, not 9. And there are severe problems with having a total of 10 DA be the essential source of truth for whole network. It would be trivial to DDoS the DAs and bring down the Tor network or at the very least, disrupt it: https://arxiv.org/abs/2509.10755.

It's the only complaint I have of the current state of Tor. Anyone should be able to run directory authority, regardless if you trust the operator or not (same as normal relays).

Anyone can. The DA code is open source and is used whenever you run a testnet. You can also run a DA on the mainnet - how do you think the 10 primary DAs exist? They're not 10 computers owned by a single organization - they're 10 mutually trusting individuals. However, most of the network won't trust you.

Couldn't you

A: Run your own network that trusts the existing plus whatever nodes you think ought to be and convince everyone that this is better if it is

B: Run a node and convince others to trust yours so that eventually there is 11 then 12 and so forth?

Interesting, thank you so much! Yeah, if those 9 really are independent entities, I’d say I don’t see many issues here.

Finding good nodes is a thorny problem for human friendship, too!

That's why the Web of Trust, or classic GNUPG key signing parties are a forgotten/ignored must have. Anyone can change and go rouge of course, but it's statistically less likely.

If I understand gp correctly, the web of trust comes after finding these human nodes, and will not help you in the process.

It doesn't work for I2P due to its design, but for things like Nostr, it works well. Essentially, the goal is to build up a list of "known" reliable relays over time, while simultaneously blacklisting anyone who joins and proves to be unreliable relying on the statistic that collaborative individuals outnumber hostile ones in any sufficiently large cohort.

Of course, it's far from being 100% effective, but it mitigates the issue significantly.

Hostile entities generally have a lot of money they can use to perform a Sybil attack.

Sure, but can't break the trusted part of the network who can remain operational in that case, even if not really anonymous anymore.

Funny and excellent comment!

> Even if there are 10x more bad nodes than good nodes [...] the good nodes should still be able to find each other

What network, distributed or decentralized, can survive such an event? Most of the protocols break down once you hit some N% threshold of the network being bad nodes, asking it to survive 1000%+ bad nodes when others usually is something like "When at least half the nodes are good". Are there existing decentralized/distributed protocols that would survive a 1000% attack of bad nodes?

I guess "predictably" is valid but what actually went wrong? After going through multiple sources I can't tell if the botnet nodes were breaking the protocol on purpose, breaking the protocol on accident, or correct implementations that nevertheless overwhelmed something.

As I understand they weren't building tunnels, so every time a legit client wanted to it has to wade through all the bad nodes to find a good one, so everything slowed right down. I was building at about 3% success rate during the issue which enables general eepsite browsing but torrenting was essentially dead

Man, I feel so out of depth with cybersecurity news.

Why does i2p (per the article) expect state sponsored attacks every February? Where are those forming from, what does the regularity achieve?

How come the operators of giant (I’m assuming illegal) botnets are available to voice their train of thought in discord?

Many state bodies involved in adversarial action have dedicated budgets for offensive cyber-warfare, credential thefts, supply chain compromises and disinformation. If they haven't used all of their budget by the end of the budget period, they'll be allocated a smaller budget for the next budget period.

Oh ffs. Whenever I think my opinion on the state of the world can’t get any lower, things somehow manage to get dumber.

I mean this is a common pattern in many large organizations, governmental and non, if you didn't use your budget it means we can save money, yayyyy! I hadn't really considered it would apply to state-backed hacking but makes sense.

State sponsored cyber attacks are news to you? It's been a thing since more than 2 decades now.

Not the attacks themselves, I would expect that kind or sabotage that actively provokes negative outcomes in people’s lives to have a more respectful/competent reasoning behind than “meh there’s a few leftovers and we had to do something”

Cool theory but that should result in other attacks that peak in February too, can you give examples?

> Why does i2p (per the article) expect state sponsored attacks every February?

Because The Invisible Internet Project (I2P) allows government dissidents to communicate without the government oversight. Censorship-resistant, peer-to-peer communication

> Where are those forming from, what does the regularity achieve?

At least PR China, Iran, Oman, Qatar, and Kuwait. censor communication between dissidents.

> How come the operators of giant (I’m assuming illegal) botnets are available to voice their train of thought in discord?

How would you identify someone as 'operators of giant botnets' before they identified themselves as 'operators of giant botnets'?

please read https://en.wikipedia.org/wiki/I2P

This answer is missing the key "regularity" part of their questions, which I would love to know more about.

That’s a great question… Currently we’re in the main Chinese holiday period with the Lunar New Year/Spring Festival/Chinese New Year, so perhaps people traveling back home from foreign lands might use the service more during this time?

I know no one using this in China. And people who can afford to travel (and have visa and passport) will have foreign sim/phone. The timing is just a coincidence

Sure, but why February and not the other 11 months?

Likely it's just a coincidence — there were other Sybil attacks that are not in February too, so the chance that you'd get 3 in Feb isn't all that low.

> The I2P development team responded by shipping version 2.11.0 just six days after the attack began.

Not wanting to be overly critical, but any net-infrastructure project kind of has to keep bot-attacks in mind and other attack vectors, in the initial design stage already. Any state-actor (and other actors, though I would assume it is often a state financing the bot network behind-the-scene) can become potentially hostile.

>hostile nodes

>they accidentally disrupted I2P while attempting to use the network as backup command-and-control infrastructure

So were they hostile or were they using it normally?

This was one of the worst writeups I ever read. Even a LinkedIn Premium post would have had more technical details, lol

This seems to be a better post about what happened, from the same site https://www.sambent.com/i2p-2-11-0-ships-post-quantum-crypto...

Those are some weird-ass visualizations. I can only assume they were AI-generated.

I'll save everyone else a click: AI slop text coupled with the strangest, most pointless visualizations I've ever seen.

Speak for yourself!

I didn’t really understand the link between Alice and Bob until I saw a green floaty dot go through a pile of spaghetti with the word compromise beneath it.

Also rewriting i2pd in Go would be the sanest step. From Java to Go is not a big challenge and you gain even more portability. Just look at Yggdrasil on how these people created meshnets running even under Android and cheap i386 netbooks.

Thus, something like this in Go should be the norm. The GC it's ideal for this, it comes with batteries charged for networking and it can be for sure be made compatible with stuff like NNCP like nothing.

It wouldn't run many times slower than i2pd in C++, it should be perfectly bearable.

This article (with high slop vibes) and another article on their site (linked in the comments) seem to suggest that post quantum encryption mitigated the Sybil attack, without explanation. I fail to understand how the two are even related.

Came here to say this. Seems like the attack was an accident and it ended. Nothing was mitigated

Definitely generated, testing a ML model I'm training. it flags as claude mostly.

> The operators admitted on Discord they accidentally disrupted I2P while attempting to use the network as backup command-and-control infrastructure ...

This is crazy to me. Discord is letting literal criminals use it's corporate services in full view to commit crimes?

Why does Discord allow a server for a botnet owner?

botnet owners dying typically come forwards and say they are trying to run a botnet, so there may be some difficulty there.

botnet owners don't typically come forwards and say they are trying to run a botnet, so there may be some difficulty in detecting them there.

Discord has a lot of terrible servers. This is one of the reasons they were not trusted when they came out and wanted to do identity verification. They already have a lot of information yet fail to do meaningful enforcement at scale.

Only a couple years ago the outrage was that Discord was too eagerly banning servers and users.

I know several people whose Discord accounts were banned because they participated in a server that later had some talk of illegal activities in one of the channels. There are similar stories all over Reddit.

If a Walmart has ~100 people in it and wants to get rid of 4 shoplifters but really sucks at selecting them well then the likely result is 4 normal people are very upset while all of the shoplifters are still there.

In the same scenario, even if Walmart is right about who they ejected 75% of the time then they still have ~1 shoplifter remaining and ~1 very upset person.

Even in an ideal world where Walmart is right about ejection 100% of the time it doesn't mean they start receiving 0 new shoplifters either, it just means the number of people wrongly made upset is 0.

Discord's problem (on both ends) lies in lack of depth in investigating bans. It takes resources to review when someone shouldn't be banned and it takes resources to make sure you ban everybody. Putting too low of resources into banning just means that both sides of the scale manage to get tipped in the wring direction at the same time.

Two things can be true at once. They can ban normal things too much and ban bad things too little.

There's servers where they just hang out, but which themselves are legitimate. Cybersecurity related ones etc. You can ban them and they'll just switch to another account within a minute. Occasionally discord or a server owner does, but everyone knows its pointless. There's probably other servers that are mostly used by cybercriminals, maybe command-and-control backups, and security researchers may stumble upon these when taking some malware apart, join them, and end up getting in contact with the owner.

In general I don't think law enforcement wants discord to take these down or ban them. These guys would have no problem to just make some IRC servers or whatever to hang out on instead, which would be much harder to surveil for law enforcement - compared to discord just forwarding them everything said by those accounts and on those servers.

Ever tried to ban a botnet owner from a service they want to use?

It’s basically impossible. They have money, IPs, identities, anything you could possibly want to evade.

They are rich in regard to the tools needed to abuse services haha.

If you just look at the messages in those kinds of discords. It's blatant. They aren't even trying to hide it.

It would be pretty funny if the age verification stuff blocked some of these folks.

Discord age verification is only for content filters, adult-themed servers, and a few other features.

They aren’t requiring age verification for everyone to join servers and chat. The headlines and panic really got away from the actual story.

I imagine because banning these things is both whack-a-mole and like finding a needle in a hay stack.

Why wouldn't they? There are Discord servers about anything you can imagine and also what you can't or don't want to image. As long as they don't start disrupting their infra Discord couldn't care less.

Also, how would you even go about classifying them as botnet operators?

A MAU is a MAU... They likely use relatively little computing capability while making numbers look really good...

I upgraded my OBSD just fine over I2P. I don't use the Java client/server, that might the reason.

I wonder how cjdns would have handled this

Isn't I2P java? The botnet uses java? I thought python or C is preferred for that kinda stuff

The official router implementation is Java. i2pd is an alternative written in C++.

Once established communication can transparently be processed through a socks proxy, or integration with SAM or similar https://i2p.net/en/docs/api/samv3/

Computers are so fast it doesn’t matter

"Since the abstraction layers have quadrupled, let's not just care about the actual performance anymore!"

Not my downvote, but which computers would that be?

More people than just myself might want one.

Communication between bots use network protocols, it doesn't matter in which language those protocols are implemented.

Is there a shittier summary anywhere, please? Or did the author reached the peak of enshittification?

Honestly, did the bot implementation have bugs or was it a proper implementation that crashed the network due to sheer numbers?

Also, how does changing the encryption standard affect anything if the bots tried to integrate correctly with the network?

Is the problem "fixed" or is it not? Elsewhere I found large number if botnet devs got pissed off with this botnet operator and 600k nodes went offline. Might this have much more to do with the situation getting better than simply changing encryption?

Also, was there any suggestion a quantum breaking attack was attempted? No. So why put the emphasis on "post quantum" in this article?

Bad. Very bad.

The video seems to be a bit more in-depth.

A bit of a tangent, but if I had a beard like that I would be making a lot more videos :)

Reply on news.ycombinator.com