A bug that certification would not have caught
On 5 November 2026, OpenSSH shipped an emergency release, 10.6p2, for a password-authentication bypass nicknamed "sshole" and tracked as CVE-2026-123158. CSO Online reported the mechanism. Musl libc's crypt() refuses passwords longer than 256 bytes and returns an asterisk as its error value. An asterisk is also what marks a password-disabled account. So a sufficiently long password can match the stored "" and log an attacker in. OX Security's analysis rates it CVSS 3.1 8.1 (High).
The flaw is narrower than the headline. Per OX and CSO, it needs OpenSSH on musl without PAM, password logins enabled, and an account with an asterisk. Glibc systems such as Debian, Ubuntu and Fedora are not affected, and neither are Dropbear or key-only configurations. Exploitation is confirmed. CSO's 5 November piece said no in-the-wild attacks had been seen at that point, but GreyNoise observed the first successful exploitation on 7 November, and CISA added the flaw to its Known Exploited Vulnerabilities catalog on 9 November. Alpine is popular in containers and is reportedly used in some embedded and router firmware. That matters for India, where those products are shipped in volume.
I also found no CERT-In advisory for this CVE. CERT-In's advisory index, last updated 17 November, listed none, though the check was not exhaustive.
The strongest case for mandatory testing
The case for a binding rule is real. Buyers of a router or smart meter cannot inspect its firmware, and a vulnerable device can be hijacked at scale. Vendors may underinvest in security because the cost of a breach falls on someone else. A mandatory baseline, tested by an accredited lab before sale, shifts that cost back to the maker. India has built one in the telecom sector.
What India's binding regime covers
The National Centre for Communication Security lists 62 published ITSARs (Indian Telecom Security Assurance Requirements) and nine designated Telecom Security Testing Laboratories. The DoT's MTCTE page defines an ITSAR as the security requirements for certification of notified telecom equipment. It does not list the covered products on the page we could read.
The Week, reporting a DoT/NCCS announcement of 30 December 2025, says ONT security certification becomes mandatory from 1 January 2026. It also says the Pro Tem scheme for IP routers and Wi-Fi CPE was extended for two years from that date, with 107 certificates issued. It reports that lab-application fees were cut by more than 50%. I could not open the DoT's primary notification, so the precise product list under the 1 January date is unconfirmed. Nothing I read extends the regime to general consumer IoT.
So the telecom regime tests defined product classes against written requirements before sale. An OpenSSH flaw disclosed on 5 November in a library interaction that has existed for years would have passed any such test taken before that date. Certification is a snapshot. A vulnerability like this one arrives afterward, and what matters then is how quickly vendors learn of it and ship a fix.
The pending framework and the proportionality problem
The Mint newspaper reported in late 2025, citing anonymous officials, that the National Security Council Secretariat is drafting a testing framework for connected devices, starting with critical sectors. I could not read the original, and I found no evidence the framework has been notified. If it arrives, its design will decide whether it helps.
Three choices matter.
- Scope. Start with devices in critical sectors, as reported, rather than every connected gadget. A blanket pre-sale mandate on all consumer IoT would burden small Indian manufacturers and startups with lab queues, and the telecom scheme's own fee cuts show how heavily costs weigh on small firms.
- Lifecycle duties over one-time tests. Require a named vulnerability-intake contact, a software bill of materials, and a stated patch window. Those duties would have helped here. An operator who knows Alpine is inside a device can check it in minutes. One who does not cannot.
- Reuse existing machinery. CERT-In's 28 April 2022 directions already require service providers, intermediaries, data centres, body corporates and government organisations to report specified incidents within 6 hours. That channel covers incidents, not vendor advisories on unexploited flaws. A short, public advisory duty for flaws in widely deployed components would fit within it without a new testing bureaucracy.
What a proportionate approach looks like
The open-source maintainers did what a good ecosystem does. They found the flaw, shipped 10.6p2 within days, and Alpine published branch-specific fixes. The weak link is downstream. Firmware vendors must notice, rebuild and push updates to devices that may be unmanaged, and regulation can address that part.
Regulators should resist rules that make shipping a patch harder. Re-certification for every firmware change would slow exactly the fixes that matter. Any framework should let a vendor push a security patch without a fresh lab cycle, and judge it afterward on disclosure and update records.
India is right to want assurance over the devices that run its networks and meters. The lesson of '*sshole' is to put the obligation where this bug was actually fixed: on prompt disclosure and patching, not on a one-time test.