Millions of Shark Robot Vacuums Are Vulnerable to Remote Hijacking — and Still Unpatched

A security researcher has found a critical, still-unpatched flaw that lets anyone with an internet connection take over an arbitrary Shark robot vacuum — watch through its camera, drive it around a stranger's home, and pull the Wi-Fi password stored on the device — using nothing more than a certificate extracted from a vacuum they legally own.

The researcher, who publishes under the handle tokay0, disclosed the vulnerability publicly on July 13 after SharkNinja, the vacuum's manufacturer, went silent following four months of back-and-forth. The company acknowledged the report in March, called it "under review" in April, promised a fix date in early July, then missed it without explanation. As of today, no patch exists and no CVE has been assigned, despite the researcher requesting one directly from MITRE after the standard 90-day disclosure window closed.

The scale is what separates this from a routine IoT bug report. By subscribing to the cloud message broker that Shark's vacuums use to receive commands, the researcher monitored traffic for 24 hours and counted more than 1.5 million distinct devices communicating with a single regional server — of which roughly 673,000, or 44%, responded in a way that confirmed they were exploitable. That's one AWS region alone; SharkNinja sells internet-connected vacuums globally.

The technical root of the problem is mundane in the way most consumer IoT failures are: a cloud access policy that was supposed to let each vacuum talk only to its own command channel instead let any device's credentials reach every other device's channel on the same server. Combined with a built-in feature that lets the vacuum's firmware execute short commands sent from the cloud — intended for legitimate maintenance and diagnostics — that misconfiguration turned a device-specific control channel into a shared one. Anyone holding valid credentials from a single purchased unit could quietly send commands to someone else entirely.

The consequences go beyond making a vacuum start or stop on command. Higher-end Shark models include an onboard camera for obstacle detection, and the researcher confirmed he could pull a live video feed and steer the motors of a second unit he'd bought specifically to test whether the flaw worked across different devices — it did. Vacuum firmware also stores a full map of the owner's home and the household Wi-Fi password in plain text, meaning a compromised vacuum can function as a foothold into the rest of a home network, not just a spying device on its own.

This is not the first time a robot vacuum's camera and cloud connection have turned against its owner. Ecovacs Deebot units were shown in 2024 to be hackable in ways that let strangers speak through the device's speaker, and a nearly identical cloud-permission flaw surfaced last year in DJI's Romo vacuum, exposing live feeds from roughly 7,000 units to a researcher who had only meant to test his own. It's the same structural failure we broke down in depth in our recent analysis of why smart home devices keep becoming hackers' easiest targets. Robot vacuums keep shipping with the same basic ingredients — a camera, a microphone or speaker, a permanent internet connection, and a full map of someone's house — and keep failing in the same basic way: a cloud permission model that assumes every device will only ever ask about itself.

Because the fix has to happen on SharkNinja's servers rather than through a firmware update owners can install themselves, there is currently nothing affected users can do except disconnect the vacuum from Wi-Fi entirely, which disables the app control and scheduling features that are the main reason people buy a connected model in the first place. SharkNinja has not issued a public statement on the vulnerability.