Researchers pull a login token out of Cloudflare's edge with a browser-era ghost bug

A Spectre-style side-channel attack ran inside one Cloudflare Worker and quietly read data from another sitting on the same machine, at speeds fast enough to lift a JSON Web Token in seconds.

ThreatVectr Newsdesk· 4 min read
A lone analyst sits in a darkened server room, face lit blue by a monitor displaying a sprawling interactive data map of cloud storage nodes and forgotten repos
Share

Key points

  • Academic researchers demonstrated a remote Spectre attack against Cloudflare Workers running in production, not a lab clone.
  • The attack leaked a JSON Web Token, the digital pass that proves who a user is, from a second Worker sharing the same server.
  • Data flowed out at up to 12 bits per second, about 360 times faster than a similar 2021 proof of concept.
  • Both the attacker code and the victim code were controlled by the research team, so no real customer data was exposed.
  • The finding revives long-standing questions about running many customers' code on shared processors.

A team of academic researchers has shown that a hardware flaw first made famous in 2018 is still very much alive at the cloud edge. Their target: Cloudflare Workers, the service that lets developers run small pieces of code inside Cloudflare's global network, close to end users.

The attack, first reported by The Hacker News, used one Worker to secretly read the memory of another Worker running on the same physical server. The prize was a JSON Web Token, or JWT, which is the small signed string a website hands your browser after you log in to prove you are still you.

If an attacker steals one, they can often impersonate the user until it expires.

What did the researchers actually do?

They built two Workers of their own. One played the attacker. One played the victim, holding a JWT in memory. Both were deployed to Cloudflare's production environment, the same infrastructure paying customers use, and the researchers waited for the platform to schedule them onto the same machine.

Once co-located, the attacker Worker ran a Spectre-style side-channel attack, meaning it tricked the processor into doing work it should not have, then measured tiny timing differences to reconstruct the secret data left behind. The leak ran at up to 12 bits per second. That is roughly 360 times the speed of a comparable 2021 experiment against Workers.

At 12 bits per second, a short JWT is a matter of seconds, not hours.

Should Cloudflare customers be worried right now?

No confirmed real-world abuse. The victim Worker was operated by the researchers, so no customer JWTs were taken. The bigger issue is architectural: Workers, like many serverless platforms, pack thousands of tenants onto shared hardware, and Spectre-class bugs live in that hardware.

In practice, the failure mode here is the one every multi-tenant cloud has been dancing around since 2018. You can isolate processes in software all you like. If two workloads share a CPU core and its caches, a patient attacker can sometimes read across the fence.

Cloudflare has previously argued that its V8 isolate model, the lightweight sandbox each Worker runs inside, plus process isolation and timing defences, made Spectre impractical at their scale. This research is a direct poke at that claim.

What ordinary users should do

Nothing dramatic. Sites you use are not known to have been breached. If you are worried about session tokens in general, the same old habits still help: log out of sensitive accounts when you are done, turn on multi-factor authentication, and be suspicious of surprise "you have been logged out" prompts.

The numbers at a glance

Detail Value
Platform targeted Cloudflare Workers (production)
Data leaked in test A JSON Web Token from a co-located Worker
Leak rate Up to 12 bits per second
Speed vs 2021 attack About 360 times faster
Real customers affected None reported; both Workers were researcher-controlled

One thing the post-mortem will say, if Cloudflare publishes one, is that shared silicon keeps writing cheques the sandbox has to cash. Serverless is cheap because it is crowded. Crowded is exactly what Spectre likes.

Operational takeaway: if your app puts long-lived bearer tokens in memory on a shared runtime, shorten their lifetime and bind them to a client fingerprint. Assume the neighbour is listening.

© 2026 Threat Vectr