Collection of potential security issues in Jellyfin This is a non exhaustive list of potential security issues found in Jellyfin. Some of these might cause controversy. Some of these are design fla…
I remember when they were arguing that you don’t need a VPN or proxy basic authentication in front of it because their team knows how to write secure code…
For those unaware, it’s a good idea to be using a service like tailscale (self hosted=headscale if you don’t want to make your login credentials tied to apple, google, or Microsoft). It’s a VPN but a lot simpler to use.
If my server is already open to everyone, what kind of potential attacks do i need to be worried a about? I dont keep personal files on my streaming server, its just videos, music and isos/roms. I dont restrict sign ups, so the idea of an unauthorized user doing something like download a video is a non issue for me really.
I do see where there could be problems for folks running jfin on the same server they keep private photos or for people who charge users for acess, but thats not me.
Am i missing something or is the main result of most of these that a “malicious” actor could dowload files jellyfin has access to without authentication?
With unrestricted signups, they can obtain their own account easily. With their own account they can enumerate all your other users.
If they have their own account they can just find your instance, make a login, collect all the proof they need that you’re hosting content you don’t own (illegally own) then serve you a court summons and ruin your life.
I wouldn’t worry about the vulnerability in the link since your already wide open. But I wouldn’t leave Jellyfin wide open either. Movie and TV studios are quite litigious.
I hope you’re at least gatekeeping behind a vpn or something.
Edit: typo
Well it’s hosted in The Netherlands and I did take some steps to protect my own identity in regards to registration info, but if the studios did take an interest i’d probably have some fun with it by decaliring bankrupcy and dragging out the appeals.
I mean, sure… but you’d actually have to reasonably liquidate most of your assets at that point. You can’t just “claim” bankruptcy and do literally nothing to sate your debts. Of course this is different on a jurisdictional basis… but overall, you have to sell a lot of your stuff in order to do a proper bankruptcy.
It can decimate any savings you have for retirement.
What assets?🤣
Fair enough if you don’t actually have any… but the courts will still make that decision for you. Some things might count that you don’t expect.
I’m not sure who needs to hear this, but unless you work as a security engineer or in another security-focused tech field, you really shouldn’t be exposing your homelab to the open internet anyway
Most people access their homelabs via VPN - i don’t see anything here that’s a problem for that use-case.
Many people host websites ;)
And I would hope those websites are extremely low-risk and not anywhere near essential infrastructure or data ;)
Can someone ELI5 this for me? I have a jellyfin docker stack set up through dockstarter and managed through portainer. I also own a domain that uses cloudflare to access my Jellyfin server. Since everything is set up through docker, the containers volumes are globally set to only have access to my media storage. Assuming that my setup is insecure, wouldn’t that just mean that “hackers” would only be able to stream free media from my server?
If you use normalized paths/file names (through *Arr stacks or docker mounts or otherwise common tools), then the hash that jellyfin sets up when it imports that media can be guessable. If someone was to go and precompile a list of hashes for content that they’re looking for at common paths that people store their files at, they can ask your server for those hashes, and if their list is sufficiently large enough to include the path that you used, your jellyfin instance WILL RESPOND WITHOUT AUTHENTICATION.
I’ve been using this example because it shows how silly this is.
In the context of Sony’s top 1000 movies, they can pre-compile the top 100 likely paths for the file (/movies, /mnt/movies, etc) then run the 100000 hash check through scripts against your instance. How long does it take to let a crawler collect http statuses on 100000 page loads? Now put that to a bot that gets jellyfin instances from a tool like shodan and add more hashes. If you flag, now onus is on you to prove you have license for content and they would have a case that you distributing (albeit weak) since your server was open to the public. This is child’s play level abuse-able. Risking that something easy like this isn’t being abused by Sony and others (you know… willing to install a rootkit on your computer types…) is a very silly stance to take.
The answer to some of this is that you can just hide the content on a more complicated and less likely to guess path. That will sufficiently change the MD5 hashes enough that you should be more or less unguessable… Instead of using
/mnt/media/movies(or /media/movies, or /movies/, etc…) make the path/mnt/k9RKiQvUwLVCjSqhb2gWTwstgKuDJx59S3J35eFzW2dgSSp84EG7PPAhf2MwCySt/media/movies. (obviously don’t use this one… use a random generator. Make your own.)The real answer should be that Jellyfin requires that all those endpoint need authorization/login. But their answer is “We don’t want to break backwards compatibility. So we won’t.” Which is a bit silly of an answer. Those who use the default installation and organize their content with *arr suites (or with default docker settings/guide settings), are most likely to have guessable MD5 hashes and are most at risk.
Edit: Oh and the other point… if the “response” against this is “well that would take too long, or be too hard. You’d need a lot of money to find all these instances and test them…”. We’re talking about the likes of Sony… The ones that installed rootkits on peoples computers for daring to put a CD into a CD-ROM drive. They’re litigious folk, and will bury you in paper and sue you to oblivion. It’s not a lot of machine time to test a single server. Setting up a couple dozen scanners and just letting it go to find content on it’s own isn’t that bad from a computational standpoint.
And another argument I’ve seen here… “Well if they hack your server then that’s illegal too, can’t make a lawsuit out of that”… Except this is normal web operations. Bots and site scanners aren’t illegal. Nor do they break any authentication mechanism (which is illegal) to do this. Specifically putting this behind authentication would make you correct. But Jellyfin didn’t do that (yet). So guess what. It’s perfectly possible for them to setup a few scanners across a few servers and do this 100% legally.
Security through obscurity isn’t security.
Edit2: Clarification on not using the path I just gave… make up your own random gibberish.
I think I understand now. Thank you! I will be changing my paths then. It’s kind of a moot point since I’ll change my paths anyway, but for the sake of my own curiosity, i have a follow up question. Feel free to disregard it if you don’t feel like taking the time to answer.
Hypothetically, my docker setup only allows jellyfin to see /mnt/user as /storage. So jellyfin would report the path to Morbius as being:
/storage/hdd1/media/movies/Morbius_all_morbed_up.mkv
when in all actuality it would be:
/mnt/user/hdd1/media/movies/Morbius_all_morbed_up.mkv
My intuition tells me that the file path that jellyfin “sees” would be the security risk. So “/storage/hdd1/…” Is that correct?
My intuition tells me that the file path that jellyfin “sees” would be the security risk.
Your intuition is correct. JF will generate the MD5 hash based on the path that it’s accessing with. So if it’s normally a unique path then you mount it into the docker container as
/movies/or/mnt/moviesor what have you… Then you lost the uniqueness, all that’s seen is the internal docker path. This is why I also lumped “using docker” into the party side by side with “using *arr stack”. Most people will find a compose file and just modify the left side of the volume declaration to point at their media. And most dockers are going to have simple internal mounts in their example compose files.Both Arr and Docker will end up pushing people to standardize the path, then the filename. Using both together compounds the issue and they tend to standardize different parts of the path.
Use a VPN
Huh, I can’t check the link right now… But if exposing Jellyfin to the Internet is not an option, then it is not ready to be shipped as the Plex replacement I have heard a lot here and on Reddit.
Put the instance behind another authentication point like a VPN or reverse proxy with SSO. That will prevent the wider Internet from accessing it without legitimate users being cut off. You should be doing this with any server you operate (like Plex), but definitely one that may have legal implications.
aaaand now you smart tv can’t connect. none of them. the clients dont even support http basic auth creds put into the URL for some crazy reason.
for advanced HTTP-level authentication you would need to run a reverse proxy on the TV’s network that would add the authentication info. for the VPN idea you would need to tunnel the TV’s network’s internet connection at the router. or set up a gateway address in the TVs network settings that would do that. or use a reverse proxy here too so that it repeats the request to the real server.
but honestly, this is the real and only secure way anyway. I wouldn’t be comfortable to expose jellyfin even if the devs are real experts. I mean vulns get discovered, in dotnet, jellyfin dependencies, linux filesystem, and reverse proxy, and honestly who has time to always tightly keep up to date with all that.
that’s not to discount the seriousness of the issue though, it’s a real shame that jellyfin is so much against security
Do we even know that Plex is better? It’s closed source and hasn’t been audited afaik
Do we even know that Plex is better? It’s closed source and hasn’t been audited afaik
Yes… because you can take the raw request your browser makes… remove your auth cookie and replay the same request and it fails.
Closed source doesn’t mean that it can’t be tested for problems. Just means that you can’t go to the code to understand why it’s a problem. You can still see that the problem exists (or doesn’t in this case).
Edit: I haven’t tested every api endpoint myself… but for video files it doesn’t work. It’s not vulnerable to the same thing that JF is in that specific case.
Agreed. I’m a bit disappointed that it’s being touted as such. If you need a local LAN option, use VLC Player.
@Scary_le_Poo I wouldn’t say never, but in most cases, you’re best served by sticking it behind wireguard- but this is also true of any service or tool you don’t intend to make available to the greater internet
So what’s the alternative? VPNs are unreliable
wireguard has been going fine here for 5+ years. only problems were when that garbage raspberry crashed as it always does (but that’s an issue with the hardware) and when the IP changes, but that’s mitigated by dynamic DNS
Unreliable how?
Possibly some ISP interference with the OpenVPN protocol. Apparently that can happen sometimes
You can always funnel all your VPN traffic through a more typical port, like 80, and there’s very little anyone can do to distinguish between your traffic and typical web traffic.
If your ISP causes issues with inbound traffic to your home network, just add another link to the chain to include a cloud-hosted server, or host it all entirely in the cloud (if you find a trustworthy one with a reasonable cost).
It’s a list from 2021 and as a cybersec researcher and Jellyfin user I didn’t see anything that would make me say “do not expose Jellyfin to the Internet”.
That’s not to say there might be something not listed, or some exploit chain using parts of this list, but at least it’s not something that has been abused over the last four years if so.
Agreed, this is a valid list of minor concerns but this is just a fearmongering post. It’s not good that some metadata can leak but if you take normal precautions (i.e. don’t run this next to your classified information storage) it’s fine to open this so your friends can watch media.
Source: me and my Masters degree in cybersecurity (but apparently OP just learned about Kerckhoff’s principle and rainbow tables in a completely incorrect context so I know how to do my job or smth lmao)
Edit: lol don’t look at OPs post history, now I know where the fearmongering came from
but if you take normal precautions (i.e. don’t run this next to your classified information storage)
oh yeah I’m pretty sure the majority of users bought a dedicated machine for Jellyfin
Source: R1 masters professor. Literally the person you would have needed to take the class from on the topic at my institution.
This is a problem simply because most paths and names will be similar due to *arr suites and docker mounts normalizing them to a standard that jellyfin wants to see. In the context of Sony’s top 1000 movies, they can pre-compile the top 100 likely paths for the file (/movies, /mnt/movies, etc) then run the 100000 hash check through scripts against your instance. How long does it take to let a crawler collect http statuses on 100000 page loads? Now put that to a bot that gets jellyfin instances from a tool like shodan and add more hashes. If you flag, now onus is on you to prove you have license for content and they would have a case that you distributing (albeit weak) since your server was open to the public. This is child’s play level abuse-able. Risking that something easy like this isn’t being abused by Sony and others (you know… willing to install a rootkit on your computer types…) is a very silly stance to take.
The hash that’s used to represent the path isn’t salted or otherwise unique.
Edit: mobile typos.
It’s nice to read something sane in these threads.
Fully agreed. There’s some stuff in the list that could leak server info or metadata about available content to the public, but the rest seems to require some knowledge before being able to exploit it, such as user IDs.
That doesn’t mean these aren’t issues, but they’re not “take your jellyfin down now” type issues either.
The last set of comments is from 2024. These have not been addressed. The fact that it is possible to stream without auth is just bonkers.
The entirity of jellyfin security is security via obscurity which is zero security at all.
“As a cybersec researcher”, the limp wristed, hand wavy approach to security should be sending up alarm bells. The fact that it doesn’t, means that likely either, you don’t take your research very seriously, or you aren’t a “cybersecurity researcher”.
“Thank you for this list. We are aware of quite a few, but for reasons of backwards compatibility they’ve never been fixed. We’d definitely like to but doing so in a non-disruptive way is the hard part.”
Is truly one of the statements of all time.
How is someone meant to guess what seems to be a randomly generated id? If they try to brute force it then you could probably set up something like fail2ban to block them after a few failed attempts.
I’m not saying video ids shouldn’t require authentication, they should but the risk of someone getting the video id seems fairly low.
It isn’t randomly generated. If you read through you would have known that.
Also, Rainbow tables.
tldr, Rainbow tables are precomputed lists of hashed values used to crack password hashes quickly. Instead of hashing each password guess on the fly, attackers use these tables to reverse hashes and find the original passwords faster, especially for weak or common ones. They’re less effective against hashes protected by a unique salt.
If the ID is the MD5 of the path, rainbow tables are completely useless. You don’t have the hash. You need to derive the hash by guessing the path to an existing file, for each file.
How unique do you suppose file system paths are?
How many hashes would one need to gather to quickly determine the root path for all files? Paths are not random so guessing the path is just a rainbow table.
The scanning for known releases becomes trivial once the file system pattern is known.
If the server is using a standard path prefix and a standard file layout and is using standard file names it isn’t that difficult to find the location of a media file and then from there it would be easier to find bore files, assuming the paths are consistent.
But even for low entropy strings, long strings are difficult to brute force, and rainbow tables are useless for this use case.
I’ve not looked but if the video id is based on its path, then surely the path includes the filename no? You can’t split a hash into its separate original parts, you either guess the entire thing or not. So in that case, the hash is going to challenging to brute force.
It’s not that challenging if you are looking for specific media files, but if you wanted to enumerate the files on a server it’s basically impossible.
You can’t say that a solution is no security at all when it requires time and intelligence to bypass.
It is at least 0.01 security.
Effort or no, if an attacker can reasonably bypass it, it’s not secure. That’s why software gets security patches all the time, why encryption/hashing algorithms can fall out of favor, and why quantum computing can be pretty fucking scary.
No system is secure.
#confidentlyincorrect
I didn’t say it’s secure, I just said it’s security.
You’re hiding behind literal definitions to avoid addressing the functional issue/implications.
This is like when somebody says “no one believes that“ and the other person finds a tweet by one person that believes the thing. The claim isn’t that literally not one person does, it’s that it’s so unusual you may as well act as if nobody does.
Surely you understand how people talk and basic vernacular?
Surely you understand how a stupid response to a silly statement like it is one of the sayings of all time can be appropriate in humorous situations, right?
I understand that you did not find it funny, but I hope that you can understand that it was my intention to be funny, and therefore a serious response is disproportionate.
I thought you were being serious as well. I’ve dealt with enough people who would genuinely make that argument so I assume nothing.
The humorous intent was not obvious.
How about 0.001 security?
So I have a NAS running Ubuntu I only keep my movies, my Jellyfin, and torrent software on in an isolated VLAN I stream from. I would think this would make any security issue with Jellyfin a dead end. I stream all content from Jellyfin domain I made and never use it locally. I stream off it at home from my VPN. This seems a safe way to stream where it can be used away from home unless I am missing something? Pointing out any holes in my logic is appreciated.
I think you can IP whitelist who can access it no? That should solve any problems
There is zero (0) chance of an attacker to know and then spoof address of your friend unless you have even bigger problems. Good filter should simply not respond to any packets making very existence of exploitable site undetectable.
Wrong use case, the expected one is friends and family watching stuff on your Jellyfin server from different homes, potentially through mobile, all with dynamic IPs
Perfect use for allowlisting based on dynamic DNS hostnames.
is that a feature in Jellyfin? and since when do all ISP subscribers have names in DNS?
deleted by creator
Does your friend have a static IP? Unlikely considering that you have to pay extra for a static IP.
We are lucky, we get two free. Technically they aren’t true static, its tied to MAC of your modem, or your router(s) – with ISP modem in bridge mode. You can pay for true static, but I have probably had the same IP for 5 years, and same with the modem/routerbeforre this one.
@Emmie @Scary_le_Poo That depends on the ISP, there’s still some out there that will give you one for free.
Who has the technical wherewithal to run Jellyfin but leaves access on the open web? I get that sharing is part of the point, but no one’s putting their media collection on an open FTP server.
The level of convenience people expect without consequences is astounding. Going to be away for home for a few days? Load stuff onto an external SSD or SD card. Phoning home remotely makes no sense.
My Jellyfin server is behind Cloudflare with IP outside of my country banned.
I got Crowdsec set up on Cloudflare, Traefik and Debian directly.
I got Jellyfin up in a docker container behind Traefik, my router opens only 80 and 443 ports and direct them to Traefik.
Jellyfin has only access to my media files which are just downloaded movies and shows hardlinked by Sonarr/Radarr from my download folder.
It is publicly exposed to be able to watch it from anywhere, and share it to family and friends.
So what? They might access the movies, even delete them, I don’t care, I’ll just hardlink them back or re-download them. What harm can they do that would justify locking everything down?
So what? They might access the movies, even delete them, I don’t care, I’ll just hardlink them back or re-download them. What harm can they do that would justify locking everything down?
Well… if “they” happen to be the rights holders or lawyers of the rights holders and they happen to enumerate their content on your system because they can guess common linux paths and likely names that their movie/show/music would appear as in your system, you’re going to care real quick when the lawsuit comes.
Friends, family using Jellyfin is the reason many have it directly available (and not behind VPN for example).
I know people are going to crucify me for this but just fucking use Plex at that point
that’s but no. I like my privacy more
And I like that my wife and kids can jump on and access my server whenever they want from any device without fuss. Everyone has their priorities! I take my privacy pretty seriously but I can’t make it the number one consideration at the cost of everything else all the time. Plus, Jellyfin is a security risk if you don’t know what you’re doing. I’m pretty tech savvy but it definitely pushes my limits so I do not feel comfortable setting it up and constantly maintaining it.
They jacked their prices, or are about to anyway. If you don’t have a lifetime Plex pass then Plex might not be a viable option. My seedbox provider has been pushing people to Jellyfin for anyone without a Plex pass.
“Jacked their prices” is a tad dramatic and if you use Plex regularly you’d be foolish not to just buy the lifetime subscription when they put it on sale for like $80 every year. The price change this year was modest except for lifetime which went from $125-$250 with a heads up meaning you could’ve still gotten it at $125 before the change.
Do you know the details of the price change?
I thought I had a lifetime Plex pass, but turns out I was on yearly and the price went up $20/year, so I bought lifetime before the price went up. My whole family uses Plex, I couldn’t handle setting up Jellyfin for everyone and their devices.
Yeah if I was just serving myself I would’ve probably stuck with Jellyfin, but my wife and kids also use my server. Because of it we pay exactly $0 a month in subscriptions. Plex lifetime pass was a very easy decision to make.
If they do a complete heel turn tomorrow and fuck us all, I could simply shut it down. The money I’ve saved so far has been worth it.
Doesn’t have a sync play feature like Jellyfin does
I understand why you might find that useful but I do not think that is exactly the most important feature in the world to most people. I could also rattle off plenty of things Plex can do that Jellyfin can’t. I have used both and the fact of the matter is just am willing to take the trade offs for the simplicity of Plex. You do you!
I’m not smart, can you tell me if having it behind a reverse proxy with certs and everything fixes any of these flaws?
short answer: no, not really
long answer, here’s an analogy that might help:
you go to
https://yourbank.comand log in with your username and password. you click the button to go to Online Bill Pay, and tell it to send ACME Plumbing $150 because they just fixed a leak under your sink.when you press “Send”, your browser does something like send a POST request to
https://yourbank.com/send-bill-paymentwith a JSON blob like{"account_id": 1234567890, "recipient": "ACME Plumbing", "amount": 150.0}(this is heavily oversimplified, no actual online bank would work like this, but it’s close enough for the analogy)and all that happens over TLS. which means it’s “secure”. but security is not an absolute, things can only be secure with a particular threat model in mind. in the case of TLS, it means that if you were doing this at a coffee shop with an open wifi connection, no one else on the coffeeshop’s wifi would be able to eavesdrop and learn your password.
(if your threat model is instead “someone at the coffeeshop looking over your shoulder while you type in your password”, no amount of TLS will save you from that)
but with the type of vulnerability Jellyfin has, someone else can simply send their own POST request to
https://yourbank.com/send-bill-paymentwith{"account_id": 1234567890, "recipient": "Bob's Shady Plumbing", "amount": 10000.0}. and your bank account will process that as you sending $10k to Bob’s Shady Plumbing.that request is also over TLS, but that doesn’t matter, because that’s security for a different level of the stack. the vulnerability is that you are logged in as account 1234567890, so you should be allowed to send those bill payment requests. random people who aren’t logged in as you should not be able to send bill payments on behalf of account 1234567890.
Many of the issues are related to unauthenticated requests. Even though your reverse proxy provides SSL, Jellyfin still won’t know the difference between you and a random internet user. So, no, your setup doesn’t mitigate the security risks much at all.
Not really, no. These are application flaws. Caddy will happily do its job and just let bad actors abuse them. (Unless you mean mTLS certs, then Caddy would only respond to those having a client certificate, which hopefully reduces the number of bad actors to your users😉)
Not unless the reverse proxy adds some layer of authentication as well. Something like HTTP basic auth, or mTLS (AKA 2-way TLS AKA client certificates)
For nginx: https://docs.nginx.com/nginx/admin-guide/security-controls/configuring-http-basic-authentication/
so if I add a user ”john” with password “mypassword” to video.example.com, you can try adding the login as: “https://john:mypassword@video.example.com”
Most HTTP clients (e.g. browsers) support adding login like that. I don’t know what other jellyfin clients do that.
The other option is to set up a VPN (I recommend wireguard)
Not unless the reverse proxy adds some layer of authentication as well.
This is correct. However I’d want to add, that this will break EVERY app-based access of jellyfin which defeats the purpose for a lot of people. Adding auth in front of jellyfin will work and allow you to use the web client only.
You kind of touched on it… I wanted to make it clear.
VPN would be a better more interoperable answer for most cases, you’ll still be stuck on devices like WebOS on LG tvs, or roku devices, or others that don’t have the ability to install vpn clients.
I’m also an absolute dumbfuck. And I can confidently tell you, as a matter of fact, that I don’t know.
I’m running SWAG reverse proxy, my DNS is not tunneled, I share my Jellyfin with others outside my network.
My primary concern is my server gets hacked, or I get charged with distributing ‘public domain movies’
Hacking, even on an insecure system, would be illegal. Any copyright troll trying to sue a single user for having a private jellyfin instance which they hacked to find out about would probably have a hard time actually making a case.
“Yeah, this one guy was distributing films to himself and a few friends. I know because I hacked him” doesn’t seem like a good case.
Nothing about this is hacking. They’re not defeating any authentication mechanism to scan your system. That’s the whole problem here. Nothing illegal about running a crawler/scanner service.
The fact that you have their content publicly accessible is not a “bad case” at all. Open FTP sites were sued plenty. It may be a bit harder to prove distribution intentions… but wouldn’t be hard to make a case that you violated copyright for the content they could enumerate.
@walden @Scary_le_Poo Only if the reverse proxy has its own login on top of Jellyfin’s, and even that only mitigates some of them.
deleted by creator
