Screen-watching implants in Ledger and Trezor wallets linked to $102 million in thefts

What happened

A supply-chain attack against Ledger and Trezor hardware wallets came into public view on 13 October, when Kim Zetter published reporting at Zero Day on implants hidden inside the devices, according to a dev.to analysis of that reporting. Hackaday followed three days later with a hardware-oriented summary. Blockchain analytics firms link roughly 315 drained accounts holding about $102 million in Bitcoin, Ethereum, Tron and other assets to the campaign, with many affected devices traced to CryptoBilis, Ledger's authorized reseller for Malaysia, Indonesia and the Philippines.

Reading the screen instead of breaking the chip

The implant never attacks the cryptography. As the dev.to post explains, it is a small microcontroller that taps the SPI lines driving the LCD. When the device displays a newly generated recovery seed, the implant decodes the framebuffer, records the 24 words and later transmits them over its own LTE modem, antenna and eSIM.

That design sidesteps the whole security model. The implant never touches the host computer or the user's WiFi network, and, as Ledger itself reportedly confirmed, it does not modify the Secure Element, the Ledger operating system or transactions. Firmware signatures stay valid and the Genuine Check passes. What is compromised is the link between the chip and the person reading the screen.

Trezor devices were hit differently. Per the same reporting, Joe Grand found implants in a Trezor One and a Model T where the microcontroller had been replaced so the wallet generated one of a dozen predetermined seeds instead of a random one. Grand notified Trezor in August, and the company's CCO said publicly on 13 October that there was "not a single piece of evidence" — the same day Zetter published Grand's findings.

Five generations of iteration

A teardown published by Mark Karpeles in September and summarized by Hackaday documents five distinct implant designs. Generation one was a rough proof of concept built with enamel wire and hot glue. Generation two laminated a transparent orange polyimide flex PCB over the main board and swapped in a smaller battery to create clearance. Generation three moved to black soldermask to blend into the board and had part numbers scraped off the modem and microcontroller to hinder forensics. Generations four and five integrated the antenna more tightly along the top edge, removed internal padding and used custom eSIM profiles.

The scale was not a one-off. Grand bought eight more Nano X units after an initial Reddit post and found implants in four of them. Separately, a user named Johannes found implants in two Nano X units bought at MediaMarkt stores in Europe, an authorized reseller with around a thousand locations across ten countries.

Warnings that preceded disclosure

Grand told Cointelegraph he alerted Ledger on 4 March 2026 about an implanted Nano X bought in Thailand in August 2025, and says the company acknowledged the finding but did not proactively warn customers. Karpeles says he reported an implant three weeks before funds were stolen and received no response.

Cheap checks before trusting a device

The dev.to author recommends non-destructive checks first: weigh the device on a 0.01g scale (a stock Nano X is 34g; implanted generation two and three units run 1.5–3g heavy), listen for a faint rattle from removed battery padding, inspect case seams for pry marks or uneven gaps, and watch for battery life 30–40 percent below specification.

A USB-C inline power meter can also help during seed generation. The author reports that an implanted device often shows an extra sustained 15–40mA draw while the recovery phrase is displayed, plus a sharp 100–250mA burst 30 seconds to five minutes afterwards as the modem wakes to upload. Generating a seed twice and observing an identical current pattern is another warning sign, since it suggests the device may not be producing real entropy. The author cautions that the meter accuracy is unverified and that none of these tests prove a device is clean — they only prove compromise cheaply, which is why they come first.

Why it matters

Hardware wallet security assumes the chip, the firmware and the signing flow are the things to defend. This campaign shows the human-display channel is equally critical: an attacker who can observe what the screen shows needs no private key, no exploit and no presence on the victim's computer or network. It also undercuts the standard advice to buy from authorized resellers, since implicated devices came through authorized channels in both Southeast Asia and Europe. And with warnings reportedly sitting with vendors months before public disclosure, the practical burden shifts to buyers to physically verify devices, and to vendors to communicate faster. The lesson is not that cold storage failed cryptographically; it is that supply-chain integrity and out-of-band exfiltration now belong squarely in the threat model.