The owner of an independent record label was about to take over a music platform he'd paid a freelancer from Fiverr to build, and his gut told him to get a second opinion first. That turned out to be a really good call, because hidden in the code was something that let an outside party run whatever they wanted on the computer it was opened on. Here's how we found it, why we started over instead of patching it, and how we got his new platform live in his own name.
The short version is that a label owner paid a Fiverr freelancer to build a music platform, and before taking it over he asked us to look at the code because something felt wrong. We found hidden code that could run commands on a computer just from opening the project in common developer tools, along with live passwords left unencrypted and no database backup. Rather than patch it, we rebuilt the platform from a clean start, put every account in his name, took it live with him and handed it over.
The owner runs an independent record label, and he wanted a platform of his own where people could stream his label's music and the artists he invited could upload theirs. He hired a freelance developer on Fiverr to build it, and the further the project went, the more his gut told him something about it wasn't right.
So before he took over the code, he did something most people skip and posted a small, fixed-fee job asking for an independent audit and a clean handoff. We reached out the day we found it and had him on the phone the same day.
At first glance it looked like any other web app. But down in one of the configuration files, below a long run of blank lines that pushed it out of sight when you scrolled, there was a chunk of code that had been scrambled on purpose so nobody could read it.
Once we unscrambled it, it turned out to be a loader, which is code whose whole job is to go get more code and run it. This one decoded itself in a few stages, checked a record on a public blockchain to find out where its instructions lived, then downloaded them and ran them quietly in the background, which meant someone outside the business could run pretty much anything they wanted on that computer. The worst part was that you didn't even have to start the app, because common developer tools read that file automatically, so just opening the project to work on it could set it off.
That wasn't the only problem, either. The project's history showed the code being added, taken out and added back again, and the owner caught that it had been put back the day before the handoff. The handoff files also had live passwords and keys for the database, sign-in, cloud storage, payments and email sitting right there unencrypted, and there was no backup of the production database at all.
We sent him two reports the next day, one written for him to make decisions from and one with the technical detail another engineer would need. We were just as careful to tell him what we couldn't know. Nobody can say who put the code there or how it got in, since developers' own computers get hit by this kind of attack and can pass it along without anyone realizing, and nobody can say whether any data was taken. We wrote more about that side of it in Hiring a freelance developer? Check the code before you take it over.
The owner already knew what he wanted, which made this part easy. He wanted to keep whatever was legitimate, move it to a fresh code repository and fresh hosting in his own name, change every password and key, and make sure there was no way back in from the old setup.
We agreed with him, so we didn't try to scrub the old project or reuse the old server at all. The first thing we handed him was a clean rebuild of the front end, about 170 files, so he could keep moving without ever opening the original code. Then we built the real platform on a clean foundation of our own and carried over the screens and layout he'd already signed off on, so the site still looked the way he wanted it to.
We kept the first version focused on music. Listeners can sign up and stream, artists join by invitation and upload their own releases, and the owner runs all of it from an admin area. Video, a shop, a community feed and a newsletter are already built into the screens, they're just switched off until he's ready for them.
Artists upload an MP3 for streaming, an optional WAV master and a square cover image, each with a size limit, and it all goes straight to cloud storage. Admins, artists and listeners each see only what they're supposed to, and the server checks who's asking on every request instead of trusting the screen. New artist pages stay hidden until the owner approves them, and he can suspend an account without losing anything or permanently delete one once he types its name to confirm.
Every song upload also asks the artist to confirm the owner's own rights wording, and the platform saves who confirmed it, when, and the exact text they agreed to, and keeps that record even if the song gets deleted later. His Terms and Conditions and Privacy Policy are posted word for word, password reset works by emailed link and signs you out everywhere once it's done, and we ran a security review before launch.
His first choice of hosting provider couldn't verify his account, so we recommended a managed host for the site and database, cloud storage for the music and artwork, and a separate service for email. He opened every one of those accounts himself, in his own name, and invited us in. The domain stayed with him the whole time, and our admin account on the platform is a normal, visible one he can remove whenever he likes.
The platform went live two weeks after we started building it, and we kept working off his feedback after that. When he asked for the next song to play on its own, we added continuous play that runs through the rest of the album and then the artist's other music. When he noticed the music cut out as he clicked around the site, we rebuilt the player so it keeps going from page to page.
At handoff we moved the code into his own GitHub account and gave him step-by-step instructions for replacing the last shared storage key himself, so the final keys to the platform are ones only he holds. We're still on as a contributor to keep it running.
From receiving the code to two written reports on what was in it
From our first look at the old code to the new platform live
Code, hosting, domain, storage and email accounts in the owner's name
| Before | After | |
|---|---|---|
| The code | Hidden loader in a configuration file that kept coming back between versions | A clean build with nothing carried over from the old project |
| Passwords and keys | Live passwords and keys left unencrypted in the handoff files | New ones created in the owner's own accounts |
| Hosting | A server the previous developer may still be able to reach | Fresh hosting, storage and email that the owner opened himself |
| Data | No backup of the production database in the handoff | A managed database in the owner's hosting account |
| Control | The owner was a guest in his own project | The owner holds every account, and our access is visible and his to remove |
This project started on Fiverr, and it's worth knowing about if you hire through Fiverr, Upwork or any marketplace like them. Plenty of the people on those sites do honest, careful work, and we're not saying otherwise. But the marketplace's job is to connect you and handle the payment, not to look at what's inside the code you get back, and you're usually hiring someone you've never met who sets up the accounts and holds the passwords until the very end.
That's where the risk comes from. If the developer's own computer has been compromised, which is one of the ways this kind of code spreads, it can end up in your project without either of you knowing until someone actually looks. So own the code repository, hosting and domain from day one and invite the freelancer in, keep the passwords and keys yourself, ask for a database backup before the final payment, and have someone with no stake in the build look at the code before you take it over.
The most important thing that happened here was the owner trusting his gut and asking someone with no stake in the original build to look before he took it over. Our part was telling him honestly what was there and what nobody could know, and then getting him to a platform he could trust without having to start from scratch on his own. If you've got a project somebody else built and you're not sure what you really have, that first look is the cheapest part of finding out.
Tell us what you're dealing with, and we'll take a look and tell you what's there, including when it all turns out fine.
Contact usUsing an AI assistant? It can send this inquiry for you. How that works.