Two researchers remotely unlock and start cars by abusing default credentials on the ForgeRock identity platform
On November 9, 2026, Shubham Shah and Sam Curry demonstrated exploit chains that took control of connected vehicles from two major automakers by chaining hardcoded credentials and WAF bypasses on the ForgeRock identity platform. For any CISO, it is a reminder that the connected car gets hacked through its identity API, not its CAN bus.
November 9, 2026. Shubham Shah, co-founder of Assetnote, and researcher Sam Curry published two write-ups describing how they took control of connected cars from two major automakers — able to start engines, flash headlights and sound horns remotely. The path does not go through the CAN bus or a vulnerable chip: it goes through the ForgeRock identity platform sitting behind the mobile apps. Default and hardcoded credentials are, once again, the flaw that opens everything.
Two automakers, two paths, one root cause
Both automakers remain anonymous — the researchers honored their request, even as enthusiasts claimed to recognize Toyota and its Lexus brand in a demo video, something Toyota would neither confirm nor deny. What is certain is that both apps rely on the same family of ForgeRock identity components (including OpenIDM) and, for the US version, the Apache Felix OSGi console.
Shah worked on the app sold in Australia; Curry on the US version. The former sums up the difference in a sentence any security professional will recognize: the US attack “required a lot more work” because the automaker “spent much more time securing the US, Canadian, and European OpenIDM environments.” In other words, the same security debt is paid differently depending on the market.
The Australian path: staging credentials in production
In the Australian version, Shah started with public data: vehicle identification numbers (VINs), found through freely available license-plate checkers. Fed into internal APIs, those VINs returned personal information and owner email addresses — the first brick of an escalation that requires no prior compromise.
The second brick is more damning: hardcoded admin credentials for ForgeRock OpenIDM, intended for the staging environment but valid in production. To use them, one obstacle remained — a WAF with heavy filtering — which Shah bypassed through path traversal flaws. He then invoked the internal admin APIs and took over target accounts knowing only their email address.
From there he could mint authentication tokens and use them to issue arbitrary commands to vehicles through the mobile app. The demonstration needs no commentary: an email address is enough to become the car’s owner.
The US path: factory defaults and a Groovy script
The US version shares the first half of the chain — WAF bypass via path traversal — but then diverges into a more involved technique of path confusion to reach the Apache Felix admin console. There, the researchers used factory default credentials on the OSGi Web Console to gain admin rights.
Once administrators of Felix, they scheduled a task that executed an inline Groovy script, giving them a shell on the identity servers. From that shell they queried the local admin APIs, which let them view and modify user account data, log into owners’ accounts, and issue the same commands as in the Australian version. The common denominator of both chains is blunt: the vehicle is never attacked directly — the identity platform is.
A two-hour patch, and a lesson that lingers
According to the disclosure timeline Curry published, the two researchers approached the automakers in early 2026, and the flaws were patched within two hours of receiving the report. That is both reassuring — these teams can respond fast — and revealing: a quick patch conceals vulnerabilities that had been exploitable since deployment and that nobody had noticed.
This is not an isolated case. Shah and Curry have a documented history on the subject — Hyundai, Kia, Subaru and others have all had their apps dismantled the same way. The common thread has not changed in years: automakers secure the vehicle as a physical object, but expose it as an API.
What this says about the state of automotive security
The connected car is a textbook case of an imported attack surface. The automaker does not build everything: it assembles an identity platform (here ForgeRock), a management console (Apache Felix), a WAF, internal APIs — and each of those components carries its own flaws. Two dominate here, and they are the oldest in the discipline.
The first is default or hardcoded credentials. A staging password that leaks into production, factory credentials never changed on an admin console — this is the class of mistake the industry has hunted for decades, and it still unlocks cars today. The second is path traversal in a WAF, a reminder that an application firewall is not a security boundary but a component to bypass, especially against motivated attackers who chain bypasses until one lands.
The operational consequence is direct for a CISO in any other sector: the security of the connected product is the security of its identity chain and its APIs. You do not fix this risk in the vehicle firmware, but in the credential lifecycle — rotation, banning staging credentials in production, removing defaults, and monitoring the internal APIs no customer should ever reach.
Verdict
What Shah and Curry demonstrated is not a spectacular technical novelty, and that is precisely what should worry us. They chained three mundane flaws — hardcoded credentials, default credentials, path traversal — and ended up starting an engine remotely.
If you operate a connected product of any kind, start with the question this case asks in reverse: which internal APIs and admin consoles are reachable behind your public facade, and with what credentials? If you are a customer of an automaker or equipment supplier, demand transparency over credential management in their identity platform — that, not the encryption of the internal bus, is where the next intrusion will happen. If you audit a ForgeRock-style identity platform, systematically test staging accounts and factory credentials in production. The rule to remember: a connected car is exactly as secure as the identity API that controls it — and an API guarded by admin:admin guards nothing.
References
- The Register — “Researchers pwn automakers’ Android apps to remotely unlock cars, honk horns” (November 9, 2026)
- Shubham Shah — “Unlocking any car from X brand in most of the world” (2026)
- Sam Curry — “ForgeRock” (2026)
- The Hacker News — “Default ‘admin:admin’ Server Password Helps Researchers Hack Two Automakers, Unlock and Start Cars” (November 9, 2026)