I hacked Gumroad’s API and broke a bunch of tools
aescreens.com
aescreens.com
I checked Gumroad's API documentation[0] and found that the product_permalink and increment_uses_count parameters weren't documented. Then I checked the Wayback Machine for the documentation at the time the exploit was discovered[1], and those parameters were documented, so the author was relying on documented, supported behavior.
Given this, Gumroad breaking all their clients with only two weeks' notice over the holidays and offering only a $500 bounty feels pretty disrespectful.
I obviously don't know about Gumroad's internals, but nothing about the original API semantics seems inherently insecure to me. Why couldn't they have just verified that the entity that owns the license key also owns the permalink? It feels like Gumroad just chose the fix that minimized work on their end while pushing a lot of work onto their API consumers and those consumers' clients.
[0] https://app.gumroad.com/api#licenses
[1] https://web.archive.org/web/20221107215637/https://app.gumro...
I guess it would be an interesting experiment to create a proxy that captures any values going out to gumroad's license verification api endpoint and change all server responses to be true instead of false. Ditto for altering the number of uses of a product in case there is a limit there too.
For anyone interested in the documentation: https://app.gumroad.com/api#licenses
Also for anyone who is new to this topic, intercepting and altering server responses can yield fascinating results in today's client side heavy code web apps.
I built that proxy server years ago and created fake responses for multiple licensing APIs provided by companies like Paddle, Setapp, MacPaw, Devmate, MacRabbit, GitTower, Gumroad, OmniGroup, among several others.
The proxy allowed me to use over 200 apps for free over the years. I tried to report this “vulnerability” (if we can call it that) to both Paddle and Setapp the same year I discovered them, but they never bothered to reply nor fix the problem. A couple of years ago, Paddle implemented a newer version of their API (v3) that uses SSL certificates and HTTP signatures to improve the security of their SDK, but I quickly found another way to bypass that protection.
To this day, even though I can easily pay for all those software licenses, I keep the proxy running because it makes it easy to visualize outbound network traffic and block unwanted HTTP requests from all the apps I use.
For example, how do you know that the setting to disable telemetry in that app you just installed actually disables telemetry? Even with a firewall like Little Snitch you can only allow/block domains/hostnames/ports but not individual API endpoints.
Thanks, but no, I did not.
At the risk of sounding like a hypocrite, I do not want to enable other people to “steal” from these companies.
These companies surely do not care if I use their apps for me, because I am just one person among several thousand potential customers. However, if I publish the code, many of these potential customers may stop purchasing their software licenses or subscriptions, and I do not want to be responsible for their loss in revenue.
Also, I spent a lot of hours reverse engineering these apps and their corresponding APIs. I would not get anything out of publishing the code: it will not get me a better job, it will not get me clean money, it will not get me good publicity, it will do no good to anyone other than the people who want to use these apps for free without doing any work.
---
> I have no idea how something like this would be built, especially in a way that it allows checking on telemetry; would be interesting to look at it.
For the sake of learning, I will give you some hints: I originally used nginx, then I wrote my own Go program, and nowadays I use envoy proxy, but pretty much any proxy server will do the trick. Then, I created and installed my own Certificate Authority to automatically trust any self-signed certificate. Maybe you can use mkcert for that. That is the easy part. The difficult part(s), and the meat of the solution, is to reverse engineer every app to understand what and how it does things. I have used several disassemblers over the years but the one I like the most is Hopper Disassembler (https://www.hopperapp.com/). You may find API keys, public SSL certificates, JSON field names, form field names, etc. and you simply need to put the pieces together to re-implement the API endpoint(s) that the app expects to use.
Sometimes it is very easy, sometimes it is very difficult. You will learn a lot for sure.
I can almost guarantee if you know about this others do too and will potentially sell or exploit it. I'd try reporting it again. Publishing it might ultimately get them to fix it, but is kind of the nuclear option.
I wrote several 'licensing systems' over the years. It's a honey trap for developers - it's so easy to overthink it and spend months on 'hardening' your licensing, to the point where you have to check the generated assembly to see if your compiler doesn't do something clever where now all of a sudden 'cracking' your software turns into replacing a single JNZ instruction with a JMP. If you really 'need' protection against this, you use a 3rd party product which essentially converts your program into something that runs on their proprietary virtual machine.
But if you do online license checking - how do you even 'fix' this? It's trivial to MITM all requests. So now you have to obfuscate your API calls, put in extra layers of crypto, hide your private keys for that will enough in your machine code, ... All of which will be cracked by someone halfway experienced in a matter of days. I think it's most likely the companies the GP reported this to just decided it's best to not bother until they have actual proof that this is a wide enough spread problem for them to make it worth putting in significant dev resources (because it does require significant resources).
I wonder how many bad actors there are or if the vast majority of apps are trustworthy.
Cert pinning can defeated, but like I said, easier said than done. Not super advanced, but still requires specialized knowledge and a willingness to put the effort in.
You don't need to create a custom proxy for that.
There are many general-purpose tools that will let you inspect/modify HTTP/HTTPS traffic between your browser and a remote server:
Wrong, you still need a custom proxy server for this to work.
The programs you suggested only help to (temporarily) inspect the Request/Response, as you say, but you would need to manually modify the responses every time they come through the proxy, which could easily translate to hundreds of requests per day while using a single app. Surely no one wants to have to sit there modifying every single HTTP request while using an app like Adobe Photoshop.
Instead, what you want is to inspect these requests/responses once using one of those programs you suggested, and then immediately translate your findings into a permanent API endpoint in your custom proxy server, which is constantly running in the background. This way you can use the app as any other user and the app will think that it is communicating with the real (remote) API.
Your suggestion to use either Burp Suite, Charles Proxy, or mitmproxy only helps if the app you are trying to crack checks for a valid license once in its entire installation lifetime. Unfortunately, the great majority of apps out there try to validate the license every few days, hours, and even minutes, for example, Sublime Text sends a request to license[.]sublimehq[.]com/check/<license> several times in a day.
The tools I mentioned all support rules for automatically modifying requests/responses. Most of them support custom scripting as well.
>Wrong, you still need a custom proxy server for this to work.
Not sure why the rudeness here. I'm just sharing some useful tools that would be easier to use than writing an entire HTTP proxy from scratch.
You'll certainly have more power writing a custom proxy from scratch, but off-the-shelf HTTP proxying tools are a good start for someone exploring this space.
Gumroad receives request from app -> Gumroad crafts response, encrypts with private key (could be Gumroad key or a dev generated key), base64 it, returns to app -> Local library decrypts with public key and then verifies response.
This of course doesn't stop someone from messing with source code and "cracking" the app, but in the case of a binary based app, that's not as straightforward.
That said, another commenter did mention that pirates are gonna pirate, so friendly easy DRM for the dev also makes sense.
On a slightly unrelated topic, I clicked through to the 'Type' page out of curiosity. The 36 second sales pitch is absolutely amazing, 0 filler content or sales speak. Makes me sad that I have absolutely no need for it.
In 2021, @Gumroad achieved: • $185.5m in creator earnings, up 30% • $10.9m in revenue, up 18%
Surely, with $10.9m revenue, a little more than $500 would've been okay to hand out for this.
25 people work at Gumroad, but none full time.
It's on their site
> $3,896,872 The amount of income earned by Gumroad digital entrepreneurs last week.
https://gumroad.com/ ~3 screens from the bottom
> Fees on Gumroad are simple: When you make a sale, we charge a 10% fee. This does not include credit card processing or PayPal fees. There are no monthly payments or other hidden charges.
That's a good business to be in!
I actually want to make a dedicated post about it cause it seems like an actually interesting utility of web3, but I also know a lot of people might just see web3 and automatically downvote.
E: the getting started page could do with being whittled down, linking someone to this and still expecting donations is ambitious! https://backed.by/pages/get-started
I think they might have kept it below $600 because of the IRS's $600 threshold* which to me is tantamount to them deciding not to pay a real bug bounty.
* https://www.bloomberg.com/news/articles/2022-12-23/irs-delay...
Besides - network access is one of the most basic functionalities of any scripting in 2023 (it was in 2013, too). How would one disallow a simple call like in the OP? I mean, they use curl to make it (even) easier to craft an HTTP call, but doing the same from straight system calls is trivial, too.
Whitehats *coffee emoji*