Most firmware changelogs are a list of things I will never notice. The one Shelly pushed on 26 October is different, because it is not about a feature at all. It is about who your relay is willing to trust, and a certificate authority in India gave that question an awkward week.
Here is the short version, as reported by GovInfoSecurity and the tink smart home blog. eMudhra, a certificate authority based in Bengaluru, revoked eight intermediate certificates after a researcher showed that a private key behind them was absurdly weak. The certificates use elliptic curve cryptography, where a key should be 48 bytes; this one was reportedly a single byte, so anyone could recover it. eMudhra says a script from April 2025 generated the bad keys, and that no certificates for real web servers were issued from them. Chrome, Firefox, Apple and Microsoft all blocked the certificates within days.
Browsers fixed this quietly, in the background, because their trust lists update centrally. Your smart plug does not work that way.
Why a light switch cares about a certificate authority
A device that makes a TLS connection needs a list of certificate authorities it trusts, and on most embedded gear that list is baked into the firmware. If an authority turns out to be compromised, nothing removes it from the device until a new firmware ships. That is the whole story of this update: Shelly says it refreshed the CA bundles and dropped the affected eMudhra roots.
Per simon42, Shelly does not claim any attack on its devices and frames this as a precaution. The firmware versions are 2.1.1 for Gen3 and Gen4, 2.0.3 for Gen2 models still in production, and 1.7.6 for the older Gen2 Plus line that is now frozen but still gets security fixes. Shelly does not publish which model gets which version, so do not go looking for a mapping table; just see what your own update entity offers. Gen1 devices run a separate firmware line and are not part of this changelog.
How worried should a local-first setup be?
Honestly, not very, and both sources I read say the same. To abuse a bad certificate an attacker still has to sit in the path of your traffic, and the exposed connections are the ones going out: Shelly Cloud, MQTT over TLS to an external broker, or HTTPS calls from scripts. A relay that only talks to Home Assistant on my own VLAN barely touches any of that.
But I like to be precise about the word barely. Plenty of local-first setups still have one or two outbound paths nobody remembers: a script that posts to a webhook, a device with the cloud toggle left on from the day it was paired, a weather call. This is a good prompt to find out which of your devices actually make outbound TLS connections. If the answer is none, you have a nice confirmation of your architecture. If the answer is a surprise, that is worth more than the update.

What I am doing about it
I do not push firmware to everything at once, and neither should you. My order is below, loosely borrowed from the tink advice to start with whatever guards a door, a gate or a camera.
- Anything with an outbound connection first: devices with the cloud enabled, MQTT over TLS, or scripts that call external HTTPS endpoints.
- Then critical switching, meaning anything on a circuit I would be annoyed to lose while I am away.
- Everything else in the following days, a few at a time, so that if an update misbehaves I know which one did it.
- Never remotely for a device I cannot physically reach. Shelly itself advises against that, and I agree.
In Home Assistant this is pleasantly boring. Each Shelly shows up as a firmware update entity under Settings, Updates, so you can install them one by one without opening the Shelly app or creating an account. Before starting, take a backup of Home Assistant itself; it costs a minute and removes the one failure mode I would actually lose sleep over.
The frozen Gen2 question
The part I find more interesting than the certificate story is the 1.7.x branch. The old Gen2 Plus devices are still getting a security fix, which is good and not something every vendor does. But they are frozen: no new features, and a successor already exists. Tink suggests considering a replacement for frozen devices that control doors or gates. I would soften that a little. A frozen device behind a firewall, doing local control, that still receives critical patches is a reasonable place to be. Replace it when it stops getting fixes, not because a newer model exists.
What I will do is write down which of my devices are on the frozen branch, so that the day the patches stop is a day I notice rather than a day I discover.
The takeaway
Trust lists age. Browsers have a machine that keeps them fresh, and your home devices have you. A compromised authority is a rare event, but the same mechanism will show up again with the next one, and the cheapest defence is the habit: know which devices phone out, and update the ones that do first.
Sources: GovInfoSecurity, tink, simon42. I have not independently tested the new firmware on my own devices yet, so treat my update order as a plan, not a report.

Leave a Reply