Randar: A Minecraft exploit that uses LLL lattice reduction to crack server RNG
github.com
github.com
Edit: the random-reversing bot was "Nostradamus" by Tim Dierks, which was declared the winner of the "supermodified" class of programs in the First International RoShamBo Programming Competition. [2]
[1] https://web.archive.org/web/20180719050311/http://webdocs.cs...
> "With his obvious technical skill, and his "cheat early and often" attitude, Tim could have a promising career as an AI programmer in the computer games industry. :)"
Instead took a path of security, authoring the TLS RFC and principal engineer in Google security. Thanks for the flashback.
I had a cool vision for “tag play” … I visualize mini RFID records on a turn table that tell Roku what to play.
I’m actually a bit relieved they have you on the team. Considering what they (Google) know about us all.
https://web.archive.org/web/20180719050236/http://webdocs.cs...
> Nostradamus was written by Tim Dierks, a VP of Engineering at Certicom, who has a lot of expertise in cryptography. The program defeats the optimal player by reverse-engineering the internal state of the random() generator, which he states "was both easier and harder than I thought it would be". To be sporting, it then plays optimally against all other opponents.
> Fork Bot was based on an idea that Dan Egnor came up with a few minutes after hearing about the contest. Since "library routines are allowed", his elegant solution was to spawn three processes with fork(), have each one make a different move, and then kill off the two that did not win. This was implemented by Andreas Junghanns in about 10 lines of code. Unfortunately, since all three moves lost to the Psychic Friends Network after the first turn, the program exited and the remainder of that match was declared forfeited.
> The Psychic Friends Network is a truly hilarious piece of obfuscated C, written by Michael Schatz and company at RST Corporation. Among other things, it uses an auxiliary function to find good karma, consults horoscopes, cooks spaghetti and (mystic) pizza to go with various kinds of fruit, #defines democrats as communists, and undefines god. We're still trying to figure out exactly what it is doing with the stack frame, but we do know that it never scores less than +998 in a match, unless it is playing against a meta-meta-cheater.
> The Matrix was written by Darse Billings, who holds the prestigious title of "Student for Life", and recently started the PhD programme at the University of Alberta. The RoShamBo program defeated every opponent with a perfect score, based on the simple principle "There is no spoon".
> Since The Matrix is also the tournament program, it has complete access to all other algorithms, data structures, and output routines, and is therefore unlikely to ever be overtaken. As a result, this category is hereby declared to be solved, and thus retired from future competitions.
I believe it works as follows: - It plays randomly for the first 998 turns (https://github.com/MrValdez/Roshambo/blob/master/rsb-iocaine...): this line is "if (*turn < trials - 2) return libra ? callback() : random() % 3;", and "libra" is initalized to (int) NULL, i.e. zero, on every invocation.
- In the last 2 turns, it uses `find_goodkarma` to comb through the stack to find where the variables that match its history and the opponents' history are stored. These the stack arrays p1hist and p2hist (https://github.com/MrValdez/Roshambo/blob/master/rsb-iocaine...)
They're easy to find because they contain 998 known values each in a ~random sequence of (0, 1, 2), and they're just upwards of the stack from the current invocation of the Psychic Friends Network.
`find_goodkarma` simply increments a pointer until the whole sequence of 998 values matches the known history.
- Then, it rewrites the history to make itself win. These lines (https://github.com/MrValdez/Roshambo/blob/master/rsb-iocaine...) never get executed, then these lines (https://github.com/MrValdez/Roshambo/blob/master/rsb-iocaine...) tally up draws so far (libra), wins (cancer) and losses (scorpio).
This line makes sure its move is the opponents' move +1 mod 3, which is the winning move: https://github.com/MrValdez/Roshambo/blob/master/rsb-iocaine...
Then, these lines repeat the same trick for the number of wins and losses. It checks whether it's p1 or p2 by comparing the addresses of the win/loss arrays, and then overwrites the wins/losses appropriately using `pizza` https://github.com/MrValdez/Roshambo/blob/master/rsb-iocaine...
in the end it returns an arbitrary value (the address of `good_hand` mod 3).
It was fun to follow but the result is kind of boring :)
amazing
There's something beautiful here and you honestly couldn't make it up.
(And the transparency got them an improvement in their security in the end.)
https://en.wikipedia.org/wiki/Hardware_random_number_generat...
In general, you can pick a random seed at the start of the day, commit to it somewhere (eg publish a hash of it on the bitcoin blockchain, or just on your website), then use that seed in a cryptographically secure PRNG on your website all day, and at the end of the day you publish the seed you already committed to.
This way people can check that you didn't cheat, but can't guess their opponents cards either.
I guess I'm not sure what publicizing their PRNG is meant to prove. It shows they didn't cheat via a very specific type of cheating but there are several other potential cheating vectors.
Yes, and this is not the only anti-cheat method they had.
I've always wondered: why aren't ADCs (e.g. mic input) and temperature sensors considered a good source of entropy, particularly if the lower bits are taken?
If you want/need to be able to argue to a third party that your entropy was good, you can spend a bit more.
How do you convince anyone that your mic input was actual real mic input, and not just a sequence of bits you made up to be convenient for you?
In a similar vein, the SciCraft minecraft server had a creeper farm which used some sort of black magic setup in order to deterministically manipulate an RNG state to trigger a "random" lightning strike at a specific block every frame in order to get better creeper drops. https://youtu.be/TM7SutJyDCk
I really don't have words for how great this post is. It made my week.
Later
A really concise explanation of the same process you can step through in Python:
https://crypto.stackexchange.com/questions/37836/problem-wit...
> which I can almost pretend to understand if I squint
This is me and all cryptography :D
This is not little payout, it sounds to me like one of the most significant exploits in anarchy minecraft history, possibly even more than nocom.
Of course, I could be underestimating this by a lot.
But that's just my opinion. The world is free to use the word however it wants.
The classic example of these is a password manager that gave out recovery codes using a PRNG. This is in that class.
The "right" solution you describe sounds overcomplicated and error-prone (now you need to think carefully about which domains are separated) compared to just using a CSPRNG.
It's not particularly. At program start-up, you seed the original PRNG. Then, you generate N numbers from the original PRNG and use those to seed N other PRNGs, then throw away the original PRNG. You don't need to think carefully about domains, you just create a new PRNG for everything you need a random number for. This makes your game easier to debug in a deterministic way, because now reproducing the behavior of one specific action that involves randomness no longer depends on every other action that involves randomness.
Even so, a non-crypto PRNG can generally compute a new random number in 2-4 ALU ops. With SIMD optimization, that can amortize to under 1 cycle per byte, which means it takes under a nanosecond to generate a new 32-bit number. I'm not sure even the best hardware-accelerated CSPRNG on modern hardware can quite say the same just yet.
But it doesn't really matter what things were like when the code was first written, it's about how it could be fixed in the present.
Now, maybe you could optimize the use of a CSPRNG here by filling large buffer(s) and sampling values from them. Some warm-up time could go a long way. However, I fear that you would run into one or more of the following problems:
- stop-the-world pause to refill the buffer (e.g. single buffer, no threading)
- synchronization delays from mutex locks (e.g. ring buffer refilled from a background thread)
- high memory usage (e.g. rotating pool of buffers, atomically swapped)
Needless to say, none of these solutions is anywhere near as simple to implement as a non-cryptographic PRNG.
Now let's consider determinism. Video games generally use a lot of differently seeded instances of the same PRNG algorithm to provide random numbers in different parts of the simulation. Since each part may demand random numbers at different rates, it's hard to replace several independent PRNGs with a single PRNG without compromising determinism. In the 4096 bytes necessary to run one instance of chacha12 at its maximum efficiency, you can fit 128 instances of xoshiro256+ or 512 instances of splitmix64 [2].
[1] = https://prng.di.unimi.it/xoshiro256plus.c
[2] = https://github.com/svaarala/duktape/blob/master/misc/splitmi...
There's nothing stopping you from having multiple CSPRNG instances, with deterministic seeds, if that's a design requirement.
So yes, a good, slightly buffered (internal state seems to be 300 bytes) implementation of a (quasi-?)CSPRNG is pretty damn close to a decent quality non-cryptographic PRNG and likely fast enough for most video games on most hardware. Though, very few people write games in Go.
I also ran it on x86 (AMD Ryzen 5600X) and got similar results; everything ran faster on a 4.6 GHz chip, but ChaCha8 stayed around 2 cpb while PCG improved a little to about 0.7 cpb.
I do have a ~10-year-old Celeron NUC laying around that I could use to test older/lower-end hardware. I can also publish the (admittedly very simple) program I'm using.
It's also worth noting that cpb is still a measure of average throughput not variance or latency and while Go's buffered ChaCha8 implementation amortizes to 2 cpb, when the 32-element buffer exhausts, generating a new number is much more expensive than the previous 31 numbers. I wouldn't say the performance is good enough for every use case.
Testing on Celeron J3455 @ 1.5 GHz (4 physical and logical cores) gave me PCG at 1.2 cpb and ChaCha8 at 2.6 cpb with cpu=1, but PCG stayed relatively constant across cpu=1,2,4,8 (worst was 1.8 cpb) while ChaCha8 slowed to 6.5 cpb at cpu=2 and 7.5 cpb at cpu=4 and cpu=8.
Back on my M1 Mac (8 logical and physical? cores), both ChaCha8 and PCG generally got better with more cores. ChaCha8 got down to 0.76 cpb at cpu=4 (then regressed a bit at cpu=8) while PCG got down to 0.26 cpb at cpu=8.
I don't think any of these results rule ChaCha8 out completely, though again I'm looking from the perspective of video games, which generally monopolize a machine while running.
ChaCha exploits instruction-level parallelism to get speed. PCG doesn't - it has a chain of instructions that must be executed sequentially. That means that when the PCG generator executes, it leaves gaps in the instruction stream that can be filled with other instructions for the game. That means a slowdown in the game that is more significant than what your benchmark suggests.
I'm going to do this and write blog about it (although I don't have a macbook), so we may be able to compare results.
I look forward to reading your blog about this.
Sharding your RNGs by player is also an option for some games, and can be easier.
The idea of a free for all bug abusing server is pretty neat, a whole ‘nother level of the game.
I guess this is what “actually fighting” (rather than just using in-game battling mechanics) would look like if the metaverse really happened ever.
because of a long history of duped high value items, PvP is just simply spamming ender crystals which deals massive damage when broken, and the defense is just how many "totems of undying" you have which absorbs lethal damage.
of course all the hacked clients automate placing ender crystals, reloading totems and identifying weak/strong locations so you're following those guidance to spam damage.
a little before that there were hacked +32,767 damage swords that will insta kill you that was patched out by the server.
What’s the difference between a weak and strong location?
I could imagine funny evolutions over time, just a random thought, if everyone is running a “glass cannon with lots of 1-hit protection” build, then I guess if players had to pick between fast/little attacks and big/slow ones, they’d favor the former. If everyone is walking around with little attacks just intended to trigger the 1-hit protections on their fellow glass cannons, then actually using the in-game armor system might turn those one hits into two-hits, making it relevant again. If the systems were properly tuned, (it could be exponentially difficult to gather 1-hit protections and extra lives, so turning those 1-hits into 2-hits could be really valuable), mechanics could be saved from obsolescence in interesting ways.
I’m not describing Minecraft at this point, just spitballing. It would be interesting to see a game designed with this evolution of “things being broken” taken into account, though. I guess that’s what Magic the Gathering is, hahaha.
IIRC ender crystals can only be placed on bedrock, and a lot of PVP happens on the bombed-out terrain that goes all the way to bedrock.
The weak terrain in the bedrock are locations where it requires multiple jumps to get out of, iirc
Balance converging around bugs and exploits is pretty typical for all PvP sandbox games with cutthroat gameplay, even if not allowed by the server. ARK: Survival Evolved and Eve Online are infamous for having huge clans (thousands of players) willing to go extreme lengths at metagaming and bug exploitation. It isn't always that rosy, ARK had certain mechanisms to dox players and their multiple Steam accounts, which I believe led to a few spillovers of the ingame relations to the real life during the Great War. Sometimes it's very basic stuff though, like building a huge tower and breaking it upon being raided, DoSing the server and crashing it, after which it rolls back to a previous backup made 10-20 min ago, making your base very hard to raid if you have active players. (an ancient thing that was fixed many years ago)
Rust (the PvP game, not the language) also had the policy of encouraging players to spread and publish bugs and exploits on YouTube, but with the different aim - so that the devs would notice and patch those faster. This resulted in a pretty robust game that is extremely hard to exploit without resorting to actual external hacks.
Sometimes you got funny situations like top guilds cheesing new raid content in their race to finish it first, leading the devs to go "cmon bro, you only played yourself" and then patching it immediately after, but almost never taking any loot away or banning anyone.
Also funny is the common situation where you don't want to _admit_ to everyone in your guild that you have at least one person running it, but you can kind of figure it out and it becomes an open secret.
1 - These are the days where you might have to clear down a dungeon 4 hours to see if a mob is up, or, even worse, to see if it's camped or not. Alternatively (or as a cover) people just parked alts but you get the idea. Also, come to think of it, Diablo had this too with maphacks and grabit and such
the fun is no longer the game in an anarchy server. It's in beating other play's strategies - which actually makes it quite close to a real war in the physical world.
For example, the 2b2t minecraft wars between factions involve logistics, misinformation and intelligence (such as spies, fakes etc). It's no longer just minecraft. But that's what makes it fun and interesting.
Welcome to the multiplayer gaming, especially games like Minecraft or ARK.
But I agree that there are far better anarchy servers than 2b around.
Isn't this basically any non-VAC CS 1.6 server?
I often find myself sharing the rng in my code for performance reasons, but stories like this definitely make me pause.
Thanks for the writeup!
If you found this amazing, take a look at this, it'll blow your mind.
https://www.youtube.com/watch?v=ea6py9q46QU and https://www.youtube.com/watch?v=GaRurhiK-Lk
At this point it feels like having PRNGs be defaults is just not that safe of a thing to offer in libraries. Like defaulting to allow TLSv1.0 or blowfish in 2024.
2b2t (an anarchy servers in genral) are Minecraft the way it is meant to be played.
Easier to have some form of land ownership/permission system for builds.
My girlfriend and I watch all the fitmc videos even though neither of us play minecraft, and love the ones detailing your insane tooling the most.
Ever since we watched the nocom one I’ve wondered what you do professionally - are you in the infosec space?
With the amount of math and computer science knowledge you put into your work I would guess more in algorithmic trading or something like that. No worries if you don’t want to answer, just curious!
Naming Baritone after Fit is actually a coincidence / joke, the repo github.com/cabaletta/baritone was the result of random brainstorming for something untaken. We only later realized it described Fit and thus added that to the readme :)