Tampered hardware wallets: the supply chain threat model that attestation does not cover

On 9 October 2026 the hardware wallet maker Ledger told a group of its own customers not to switch on the device they had just bought. Those customers had one thing in common. They had all bought from CryptoBilis, a Malaysia based retailer that sold Ledger devices across South East Asia. Three days later Ledger said it had found an unauthorised hardware implant inside one of the affected devices. According to Ledger's investigation update, that implant read the 24 word recovery phrase off the device screen during setup and transmitted it over a mobile network. No part of Ledger's cryptography had been broken. The device had been opened and modified before the buyer ever saw it.

That is a different problem from the one most self custody guidance addresses. The standard advice assumes the device is honest and the owner is the weak point, so it concentrates on where you write the seed phrase down, whether you ever type it into a phone, and how you survive a house fire. A device that was tampered with in the distribution chain breaks the assumption underneath all of that advice. This article sets out what has been verified, by whom, and on which date, what a buyer can realistically check, and where the evidence remains thin. It is a factual reference and it does not constitute investment or financial advice.

What happened to the Ledger devices sold through CryptoBilis?

Ledger has confirmed that at least one device sold by the reseller CryptoBilis contained an unauthorised hardware implant, and that the implant captured the recovery phrase shown on the screen. Ledger's support account stated on 10 October 2026 that "Ledger can confirm that one of the impacted users' devices contained an unauthorized hardware implant", as reported by CoinDesk and BleepingComputer. In its 13 October update Ledger described how the implant worked. When the user set the device up and the screen displayed the recovery phrase, a microcontroller inside the implant recognised and decoded the words, then sent them out over LTE using a provisioned eSIM.

Ledger was specific about what the implant did not touch. It "did not modify the device's secure Element, the chip designed to protect private keys, or its operating system", and it "did not alter transactions or signatures", per the statement reported by CoinDesk on 13 October 2026. In other words the secure chip kept doing its job correctly while a second, unrelated piece of hardware watched the screen.

The teardown detail comes from outside Ledger. BleepingComputer, reporting on 12 October 2026 and citing security researcher Mark Karpelès, described an implant built around an ST STM32G431KBU6 microcontroller, a two millimetre by two millimetre eSIM in an eight pad package, a shielded LTE module, an antenna wire routed along the inside edge of the case, and a smaller replacement battery that powered both the implant and the wallet. Two physical variants were described, one with components tucked under the battery and a neater one that fitted into the foam padding beneath the OLED display. That same report says researchers first documented an implant in a Ledger device on 16 September 2026, which is three weeks before Ledger opened its public investigation.

Timeline of the CryptoBilis hardware implant incident, with the source that reported each step

Date	Event	Reported by
16 September 2026	Researchers first document an implant inside a Ledger device	BleepingComputer, 12 October 2026
10 September 2026	Earliest of the two Singapore cases later reported to police	Singapore Police Force advisory, 28 October 2026
9 October 2026	Ledger opens an investigation, asks CryptoBilis to pause all sales and shipments, tells buyers of the previous 90 days not to initialise their devices	CoinDesk, 9 October 2026
9 October 2026	Trezor stops selling through CryptoBilis, before any indication it was itself affected	Trezor blog, 30 October 2026
10 October 2026	Ledger confirms an unauthorised hardware implant in one impacted user's device	Ledger support statement, via CoinDesk and BleepingComputer
13 October 2026	Ledger describes the screen reading and LTE exfiltration mechanism. MediaMarkt suspends Ledger sales in store and online after a German customer reports a tampered unit. Ledger asks retailers in seven Asian markets to suspend sales	CoinDesk and The Block, 13 October 2026. FinanceFeeds, 29 October 2026
16 October 2026	Bitquery's running on chain tally reaches 102.0 million dollars across 335 victim wallets on six chains	Bitquery investigation page, last updated 16 October 2026 at 14:40 UTC
28 October 2026	Singapore Police Force issues a public advisory on tampered hardware wallets bought from retail stores in Singapore	police.gov.sg, 28 October 2026
30 October 2026	Trezor publishes a report of a suspected implant in a Safe 5 bought from CryptoBilis in August 2026	Trezor blog, 30 October 2026

The geography widened quickly. Ledger said on 13 October 2026 that it had asked retailers in Indonesia, Singapore, Thailand, the Philippines, Vietnam, Malaysia and Hong Kong to suspend sales while the investigation continued, according to FinanceFeeds on 29 October 2026. BleepingComputer also reported resales reaching Japan through Amazon Japan and Yahoo Auctions. On 13 October 2026 the European electronics chain MediaMarkt pulled Ledger products from its shelves and its website after a German customer reported a tampered unit, telling The Block that it took the incident "very seriously".

How much was actually stolen, and who says so?

Nobody authoritative has confirmed a total, and the published figures range from 86 million to 102 million dollars depending on which researcher counted and on which day. Ledger itself has declined to endorse any of them. The Block quoted the company saying that "Figures circulating online come from third parties and Ledger has not fully verified them", and CoinDesk reported on 13 October that Ledger "has not confirmed the total number of victims or the value stolen".

Published loss estimates for the CryptoBilis incident, with the basis and date of each

Figure	Who produced it	Date	Basis	Confirmed by the vendor
86 million dollars	Pseudonymous on chain investigator using the handle Specter, relayed by CoinDesk	9 October 2026	Suspected theft addresses on Bitcoin, Ethereum and Tron	No
Up to 93.2 million dollars across more than 300 wallets	On chain researchers, relayed by CoinDesk	13 October 2026	Address clustering across multiple chains	No
102.0 million dollars across 335 victim wallets	Bitquery, a blockchain data firm	Last updated 16 October 2026, 14:40 UTC	Tracing across six chains, following transfers through swap services and bridges, matching transaction timing and amounts, with a second internal rebuild from raw data	No
About S$3.5 million, roughly 2.7 million US dollars	Singapore Police Force	28 October 2026	At least two reported cases since 10 September 2026	Police said they were in contact with Ledger, which was reviewing the cases

Bitquery's chain level split, as of its 16 October update, put 79.3 million dollars of the total on Tron across 151 victim wallets, 16.8 million on Bitcoin across 122 wallets, 3.7 million on Ethereum across 33, 1.45 million on BNB Chain across 27, 0.58 million on Polygon across 13 and 0.25 million on Solana across four. The same page reported that roughly 81 million dollars was still traceable on chain on 16 October, down from about 97 million on 10 October, and that Tether had frozen 10.0 million dollars of USDT across 20 wallets. Issuer level freezing is one of the few recovery levers that exists here, and it only works on centrally issued tokens, which is a point worth holding next to any discussion of stablecoin regulation and US Treasury guidelines.

Treat all of these totals as estimates built from address attribution rather than from confirmed victim counts. Address clustering can over count by sweeping in unrelated wallets and under count by missing victims who never reported anything. The loss totals are the weakest numbers in this article, and the only figure here with an institutional author is the Singapore police total of about S$3.5 million across two cases.

Why is a tampered device a different threat model from a lost seed phrase?

Because the standard seed phrase threat model assumes the hardware is honest, and supply chain tampering removes that assumption before you have made a single decision. In the familiar model your private keys never leave the secure chip, the screen tells you the truth about what you are signing, and your remaining jobs are to keep the recovery phrase offline and to resist phishing. Every one of those protections routes through the device. If an attacker has already been inside the case, the protections are not wrong, they are simply bypassed.

The implant described by Ledger did not need to defeat the secure element at all. It needed only to be in the room when the screen showed the recovery phrase once, during setup, which is the one moment the phrase is guaranteed to be displayed in full. After that the attacker holds the master secret and can drain the wallet at any later time from anywhere, with no further access to the device and no interaction with the owner. That is why some victims noticed nothing until funds moved.

The two models also differ in who can fix them. Seed phrase hygiene is something you control. Supply chain integrity is something a manufacturer, a distributor and a shop control on your behalf, and you inherit the result. In this threat model the device in your hand is the adversary, and no amount of careful seed storage changes that. It is the same structural point that runs through debates about bank custody of bitcoin under FDIC approval and about Basel III capital requirements for crypto. Self custody does not remove counterparty risk, it relocates it from a custodian to a hardware supply chain, and the second one is harder for an ordinary buyer to audit than the first.

For scale, Chainalysis reported on 18 December 2025 that it had traced over 3.4 billion dollars of crypto theft from January to early December 2025, of which 713 million dollars came from personal wallet compromises, across roughly 158,000 incidents affecting at least 80,000 unique victims. Personal wallets accounted for 20 per cent of all value stolen in 2025, down from 44 per cent in 2024. Chainalysis itself cautions that "the true number of compromises is likely far greater" and disclaims warranties on the accuracy and completeness of its figures. These are commercial on chain analytics with a proprietary method rather than audited statistics, and they do not isolate hardware tampering as a category. Readers who want a government compiled complaint series can go to the United States Internet Crime Complaint Center annual reports, bearing in mind that complaint data is self reported and skewed to one country.

Does a genuine device check prove that nobody added hardware?

No, and the manufacturers say so themselves. Ledger's own buying guidance, dated 6 March 2026 and updated 18 September 2026, explains that the Genuine Check works by sending a challenge to the device, to which the secure element replies with a cryptographic signature using a private key injected at the factory, verified against Ledger's servers. The same page then states that the check "cannot detect unauthorized physical modifications to the hardware, such as spying implants, if the original Secure Element remains intact". That is exactly the condition the CryptoBilis implant satisfied. Attestation proves that a genuine secure chip is present, not that the device around that chip is unmodified.

Trezor's equivalent works in a comparable way. According to Trezor's documentation, Trezor Suite sends a random challenge, the Optiga Trust M secure element signs it, and the signature together with a certificate written during production authenticates that particular secure element as one Trezor provisioned. The documentation describes this as making a device "significantly harder to tamper with" and supports it on the Safe 3, Safe 5 and Safe 7. It does not claim to detect an extra circuit board, and the Safe 5 report of 30 October 2026 involved precisely that, an added board sitting between the main board and the screen.

This limitation is not new and it is not a surprise to researchers. A peer reviewed workshop paper presented at the ACM CCS Workshop on Decentralized Finance and Security in 2021, by Dabrowski, Pfeffer, Reichel, Mai, Weippl and Franz, argued that hardware wallets "can be replaced or manipulated in a number of hard-to-detect ways pre- or post-delivery to the user" and that "implemented (remote) attestation and authenticity checks fail their purpose for multiple reasons". The authors concluded that tightened attestation and encrypted communications would not fix the underlying architecture, and proposed collaborative key generation and signing so that neither a compromised wallet nor a compromised computer could unilaterally manipulate keys. That paper is five years old and predates the specific devices discussed here, which weakens it as evidence about any current model, but it does establish that the gap was documented long before October 2026.

What each verification mechanism proves, and what it does not

Mechanism	What it proves	What it does not prove	Source and tier
Ledger Genuine Check	A genuine Ledger provisioned secure element is present and answers a cryptographic challenge	That nothing was added to the board, screen, buttons, battery compartment or radio paths	Ledger Academy guidance, updated 18 September 2026. Manufacturer page, not an independent source
Trezor Safe device authentication check	That this specific Optiga Trust M secure element was configured on Trezor's production line	That no additional circuit board, cellular module or antenna is present inside the case	Trezor documentation. Manufacturer page, not an independent source
Coldcard numbered tamper evident bag	That the number on the bag exterior, the number on the tear off tab inside, and the number the device shows on screen all agree	That a sufficiently resourced attacker could not reproduce or re-bag packaging. Coinkite states the checks "do not replace buying from a known source"	Coldcard security and verification documentation. Manufacturer page, not an independent source
Boot time flash checksum, shown as GENUINE or CAUTION	That the flash contents match the checksum the secure element recorded	That the firmware checked is the only code running, or that no separate component is reading the display	Coldcard documentation. Manufacturer page, not an independent source
User supplied entropy at seed generation, such as 50 dice rolls or 128 coin flips	That the seed does not come solely from the device, which defeats a pre-determined seed implant	Nothing about an implant that reads the finished seed off the screen afterwards	Coldcard setup documentation. Manufacturer page, not an independent source
Your own physical inspection, by weight or by opening the case	Very little that is reliable	Trezor's position, reported 2 November 2026, is that "there is no dependable way for users to inspect a device on their own" and that small weight differences have other causes	Help Net Security, 2 November 2026, reporting Trezor's advice

Which hardware wallet tampering cases are actually documented?

Several, going back to 2018, and they split into three mechanisms rather than one. The first is a device that generates a seed the attacker already knows. The second is a device whose integrity check is satisfied by a copy of legitimate firmware stored alongside the malicious code. The third, which is what October 2026 introduced at scale, is an added component that simply reads the screen and radios the result out.

In March 2018 the researcher Saleem Rashid published an analysis of the Ledger Nano S showing that the device paired a non secure microcontroller, an STM32F042K6, with a secure element, an ST31H320, and that the microcontroller's firmware could not be properly verified. A tampered microcontroller could hook the display functions and fix the recovery seed at setup so that every derived private key was predictable to the attacker. Rashid noted that the secure element's attestation could vouch for itself but could not detect a compromised microcontroller, and that Ledger had argued its attestation was strong enough that "Ledger wallets do not require tamper-resistant packaging". Ledger shipped a Nano S firmware update on 6 March 2018, before the write up was published, while the fix for the Ledger Blue was still unreleased at the time of writing.

In May 2023 Kaspersky published a teardown of a counterfeit Trezor Model T that had already been used to steal a victim's funds. The fake used an STM32F429 with flash read out protection fully disabled, where the genuine Model T uses an STM32F427 at protection level RDP 2, and the substituted chip showed solder traces. The modified firmware removed bootloader checks and signature verification, and at initialisation or reset it replaced the randomly generated seed with one of 20 seed phrases baked into the firmware. Passphrase protection was cut down to the first character only, which together with the no passphrase option left 64 possibilities, so an attacker needed to try 64 times 20, or 1,280 variants, to open a given fake. The tells were a bootloader version 2.0.4 that the vendor had never released, and a case closed with glue and tape instead of ultrasonic bonding. Kaspersky's point was that to the victim the device looked exactly like a real one.

On 13 October 2026 the security journalist Kim Zetter reported that the hardware hacker Joe Grand had examined two compromised Trezor devices. A Model T bought through Facebook Marketplace in Canada in 2025 had a replaced microcontroller with modified firmware plus an external microSD card soldered to internal connectors, holding legitimate firmware so the integrity check would pass. A Trezor One obtained from a TikTok shop in August 2026 used a microcontroller with twice the required internal flash, half malicious code and half official firmware kept for the same purpose. Both produced pre-determined seeds. Zetter reported that Grand found Trezor Suite did flag these on the latest version but still allowed the devices to be used, and that Trezor's claim its integrity checks would catch such devices was one Grand contested. The Trezor One owner had lost about 50 dollars before noticing.

Packaging has been forged too. CoinDesk reported on 17 June 2021 that scammers were posting fake Ledger Nano devices to people whose home addresses had been exposed in Ledger's 2020 e-commerce database breach, which compromised about 272,000 customer records. The parcels contained a shrink wrapped box, a sealed bag bearing Ledger's logo, a letter claiming the wallet had to be replaced, and a flash drive soldered inside the device running a counterfeit app designed to harvest the 24 words. Ledger's then chief information security officer Matt Johnson advised people to "be suspicious of receiving a free product in the mail that you didn't order".

Documented hardware wallet tampering cases, 2018 to 2026, by mechanism and channel

Case and date	Mechanism	Purchase channel	Evidence
Ledger Nano S, March 2018	Replaced non secure microcontroller firmware fixes the seed at setup	Demonstrated as a reseller attack, not an observed theft	Published proof of concept by Saleem Rashid, patched firmware from Ledger
Counterfeit Trezor Model T, May 2023	Substituted microcontroller, signature checks removed, seed drawn from 20 pre-generated phrases	Third party marketplace seller described as trusted	Kaspersky laboratory teardown of a device used in a real theft
Trezor Model T and Trezor One, 2025 and August 2026	Replaced microcontroller plus a stored copy of legitimate firmware to satisfy integrity checks, pre-determined seeds	Facebook Marketplace in Canada, and a TikTok shop	Teardown by Joe Grand, reported by Kim Zetter on 13 October 2026
Ledger devices via CryptoBilis, October 2026	Added implant with microcontroller, eSIM and LTE module reads the recovery phrase off the screen	An authorised reseller, plus onward resale in other markets	Ledger confirmed an implant in one device. Independent teardown detail from Mark Karpelès via BleepingComputer
Trezor Safe 5 via CryptoBilis, reported 28 October 2026	Extra circuit board between main board and screen with a cellular module and antenna	CryptoBilis via Tokopedia, bought August 2026	Single customer report and video interview. Trezor had not received the device as of 30 October 2026
Counterfeit Ledger devices posted to breach victims, 2021	Fake device with soldered flash drive and counterfeit app to harvest the 24 words	Unsolicited post, addresses from Ledger's 2020 data breach	CoinDesk reporting, 17 June 2021, with a statement from Ledger's CISO

Across every case above the compromised device reached its owner through a channel other than the manufacturer's own shop, which is the single clearest pattern in the public record. That pattern is suggestive rather than conclusive. Nobody has published a denominator, so we cannot say what share of direct purchases were checked and found clean, and devices bought direct are also less likely to be examined and reported in the first place. Ledger has said it has had no reports concerning devices bought directly from it, which is a company statement about absence of evidence, not an audit.

What can a buyer verify, and what can nobody verify?

You can verify provenance, packaging consistency, firmware authenticity and the origin of your own seed entropy. You cannot verify, on your own, that nothing has been added inside the case. The most striking thing in this episode is that Trezor told its customers not to try. Help Net Security reported on 2 November 2026 that Trezor offered replacement devices or vouchers and specifically advised buyers not to weigh or open their wallets looking for implants, because "there is no dependable way for users to inspect a device on their own" and small differences in weight can have other causes.

The research literature supports that pessimism. Detecting added or modified hardware at chip scale is an open problem in its own field, and the non destructive techniques under development need laboratory equipment. Andrew Huang's 2023 preprint on Infra-Red In-Situ inspection, arXiv 2303.07406, which is a preprint and has not been peer reviewed, proposes short wave infrared imaging through the back of a chip and frames the contribution as hardening self test methods "by ruling out the existence of extra circuitry that can hide a hardware trojan with a test bypass". A method that requires an infrared imaging rig is not a check a buyer performs on a kitchen table.

What remains in the buyer's hands is a short list, and it is worth being honest that it is mostly about choosing a channel rather than inspecting a device.

• Buy from the manufacturer's own shop, or from a reseller named on the manufacturer's current list, and check that list on the day you buy rather than relying on a screenshot or a memory. CryptoBilis was an authorised reseller at the time, so this control reduces exposure rather than eliminating it.
• Treat marketplaces, auction sites, social commerce and unsolicited parcels as out of scope. Three of the documented cases came from Facebook Marketplace, a TikTok shop and the post.
• Run the manufacturer's authenticity check on first connection, and never opt out of it. Trezor's documentation allows users to skip the device authentication check and advises against doing so.
• Where the device supports it, add your own entropy at seed generation. Coldcard documents generating a seed from device entropy plus 65 keypresses, 50 dice rolls or 128 coin flips, which defeats the pre-determined seed class of attack even if it does nothing about a screen reading implant.
• Where the device ships in numbered tamper evident packaging, check that all three numbers agree, meaning the bag exterior, the tear off tab inside and the number the device displays. Coinkite's own documentation says these checks "do not replace buying from a known source or checking the device at first use".
• Do not treat intact shrink wrap or a holographic seal as evidence. Both have been forged, in the 2021 posted fake Ledger parcels and in the 2023 counterfeit Model T.

Choosing the purchase channel is the highest value control an ordinary buyer still holds, because every other check either verifies a chip rather than a device or has already been defeated in a documented case. That is an uncomfortable conclusion for self custody advocacy, and it should be stated as such rather than dressed up. The counter argument is real too, which is that the alternative is trusting an intermediary with your assets, and the record of intermediaries is not spotless either, as the pattern described in the surge in fintech bankruptcies and the custody questions around social trading platforms in 2026 both illustrate. The honest position is that both routes carry risk you cannot fully inspect, and the risks are different in kind.

How strong is the evidence behind each claim in this article?

Unevenly, and the loss figures are the softest part. The table below scores each substantive claim made above. It is written to be used against the article.

Evidence strength of each claim in this article, graded against the source actually available

Claim	Type of evidence	Strength
An unauthorised hardware implant was found in at least one Ledger device sold by CryptoBilis	Manufacturer statement, reported consistently by several outlets	Strong for one device. Ledger has not published a device count, so the scope is unknown
The implant read the recovery phrase from the screen and sent it over LTE via an eSIM	Manufacturer statement plus a named researcher's teardown relayed by a security outlet	Strong on mechanism. No full technical write up has been published, and Ledger has not confirmed the component list
Losses of 86 million to 102 million dollars	Third party on chain address attribution, one source pseudonymous, vendor explicitly unconfirmed	Weak. Range is wide, method is not reproducible by the reader, and no victim count underpins it
Two Singapore cases with about S$3.5 million in losses since 10 September 2026	Government advisory from the Singapore Police Force	Strong for the two cases and the figure. The advisory does not attribute the tampering to any named reseller
A Trezor Safe 5 sold by CryptoBilis contained a similar implant	One customer report plus a video interview, device not yet examined by Trezor as of 30 October 2026	Weak. Trezor calls it credible but unconfirmed, and says it will not compare the two cases further
Genuine and authenticity checks cannot detect an added implant	Manufacturer documentation from Ledger stating the limitation directly	Strong on the specific point, but it rests on a commercial page rather than an independent audit
Attestation designs fail for structural reasons, not only by accident	Peer reviewed workshop paper, ACM CCS DeFi workshop 2021	Moderate. Peer reviewed and directly on point, but published five years ago and not about the current models
Owners cannot reliably inspect their own devices for implants	Trezor's own advice to customers plus the state of the hardware inspection literature, including one preprint	Moderate. Consistent across sources, but it is an argument from absence of a practical method rather than a measured failure rate
Buying direct from the manufacturer materially reduces this risk	Pattern across documented cases plus manufacturer advice. No denominator and no controlled comparison	Weak to moderate. Plausible and consistent, but vulnerable to reporting bias in both directions
Personal wallet compromises were 20 per cent of value stolen in 2025, down from 44 per cent in 2024	Commercial blockchain analytics with a proprietary method and a published accuracy disclaimer	Moderate for direction of travel, weak for precision. Does not isolate hardware tampering
EU Cyber Resilience Act dates and obligations	European Commission policy page for Regulation (EU) 2024/2847	Strong on dates and scope
No current rule requires makers to make reseller tampering detectable by the buyer	Our reading of the published scope of the Cyber Resilience Act. We found no such provision, which is not the same as proving none exists	Weak. This is a reading, not a legal opinion, and it should be checked against the final implementing standards

Does any regulation reach reseller tampering?

Partly, and the part that exists is about vulnerability reporting rather than about physical tamper evidence. The European Union's Cyber Resilience Act, Regulation (EU) 2024/2847, covers products with digital elements, which includes connectable hardware, and it requires manufacturers to meet cybersecurity requirements across the product lifecycle and to manage vulnerabilities. According to the European Commission, the regulation entered into force on 10 December 2024, the reporting obligations for actively exploited vulnerabilities began on 11 September 2026, and the main obligations apply from 11 December 2027. Certain higher risk products need third party conformity assessment, and compliant products carry CE marking enforced by national market surveillance authorities.

Whether an implant soldered in by someone in the distribution chain is a vulnerability in the manufacturer's product is a genuinely open question, and we are not in a position to answer it. The device as designed behaved correctly. The failure sits in channel control, which is a commercial and logistical function rather than a software one. Nothing we found in the published scope of the Cyber Resilience Act obliges a maker to give buyers a means of detecting post-manufacture physical modification, and that gap is the one this incident exposed. Readers should treat that as our reading of a regulation whose implementing standards are still being written, not as settled law.

What should a holder do now?

If you bought a device from an affected channel, act on the assumption that the recovery phrase is already in someone else's hands. Both manufacturers gave the same instruction. Ledger told buyers who had not yet set up a device not to initialise it, and told those who had to move assets to a new device with a newly generated recovery phrase. Trezor told everyone who bought through CryptoBilis from March 2026 onwards to move funds to a new wallet with a new seed as soon as possible, and added that if no other secure hardware wallet is available, moving to an exchange or a software wallet is preferable to leaving funds where they are.

Treat any recovery phrase that a suspect device has ever displayed as public, because a new passphrase or a new PIN on the same seed does not help once the words have left the room. A new seed on a device you trust is the only remedy that addresses the actual exposure.

Ledger also asked anyone who lost funds to "coordinate with law enforcement before opening the signer", since an intact device is evidence. The Singapore Police Force advisory of 28 October 2026 gives three precautions, which are to follow official advisories from your wallet provider, to stop entering recovery phrases or transacting on a device you suspect and contact the provider through official channels, and never to disclose private keys or recovery phrases to anyone, explicitly including people claiming to represent wallet providers, exchanges or law enforcement agencies. The force asks victims to report through the police hotline on 1800-255-0000 or its i-Witness channel.

The durable lesson is narrower than the headline figures suggest, which is that a passing genuine check tells you a chip is authentic and tells you nothing about what else is inside the case. None of the above is investment or financial advice, and it is not legal advice either. It is a description of what has been published, by whom, and on what date.

Frequently asked questions

Is my Ledger or Trezor unsafe if I bought it directly from the manufacturer?

No confirmed case in this incident involves a device bought directly from Ledger, and Ledger has said it has had no reports of such devices. That is a company statement rather than an audit, and absence of reports is not the same as absence of tampering. Every documented compromise so far arrived through a reseller, a marketplace or the post.

Can I tell whether my device has an implant by opening it or weighing it?

Trezor's advice, reported on 2 November 2026, is explicitly not to try, on the grounds that there is no dependable way for a user to inspect a device alone and that small weight differences have other causes. Opening the case also destroys evidence, and Ledger asked affected users to coordinate with law enforcement before opening a device. Send a suspect unit to the manufacturer instead.

How much was stolen in the CryptoBilis incident?

No confirmed total exists. Published estimates run from 86 million dollars, relayed by CoinDesk on 9 October 2026 from a pseudonymous on chain investigator, to 102.0 million dollars across 335 wallets in Bitquery's tally last updated on 16 October 2026. Ledger has stated that circulating figures come from third parties and that it has not fully verified them. The only figure from an institution is the Singapore police total of about S$3.5 million across two cases.

Did the attackers break the secure element or the cryptography?

No. Ledger stated that the implant did not modify the secure element or the operating system and did not alter transactions or signatures. The implant read the recovery phrase from the screen during setup, which is a different attack surface entirely, and it is the reason the device's own integrity checks reported nothing wrong.

Does a passphrase or a PIN protect me from this attack?

Not reliably. A PIN protects a device you physically hold, and an attacker holding your recovery phrase does not need your device. A passphrase on top of the seed can help in principle, but the 2023 Kaspersky teardown showed passphrase handling can itself be sabotaged in tampered firmware, reduced in that case to the first character and 64 possibilities. Assume a displayed seed is lost and start again on hardware you trust.

What is the single most useful habit to adopt?

Buy direct, verify authenticity on first connection, and supply your own entropy at seed generation if the device supports it. None of the three is sufficient alone. Together they close the pre-determined seed class of attack and remove the reseller from your trust chain, which is where every documented failure has occurred so far.

Sources

• CoinDesk, 9 October 2026, Ledger investigates potential wallet tampering after reports of 86 million dollars in crypto stolen
• BleepingComputer, 12 October 2026, Ledger confirms hardware wallet spy implant behind crypto thefts
• CoinDesk, 13 October 2026, Ledger finds covert cellular implant stole hardware wallet recovery phrases from screens
• The Block, 13 October 2026, Ledger confirms hardware implant in CryptoBilis reseller units as MediaMarkt suspends sales in Europe
• Singapore Police Force, 28 October 2026, Police advisory on tampered cryptocurrency hardware wallets
• Trezor, 30 October 2026, Implant reported in a Trezor sold by CryptoBilis
• Help Net Security, 2 November 2026, Trezor warns some of its wallets may hide a seed-stealing implant
• Kim Zetter, Zero Day, 13 October 2026, Trezor wallets also compromised with implants
• Bitquery, investigation page last updated 16 October 2026, Ledger CryptoBilis hack, 102.0 million dollars drained
• FinanceFeeds, 29 October 2026, Tampered Ledger wallets cost Singapore users
• Ledger Academy, updated 18 September 2026, Best practices to securely buy your Ledger signer
• Trezor, Trezor Safe device authentication check
• Coinkite, Coldcard security and verification
• Saleem Rashid, 20 March 2018, Breaking the Ledger security model
• Kaspersky, 10 May 2023, Review and analysis of a fake Trezor cryptowallet
• CoinDesk, 17 June 2021, Scammers are sending Ledger users fake hardware wallets
• Dabrowski, Pfeffer, Reichel, Mai, Weippl and Franz, Better keep cash in your boots, hardware wallets are the new single point of failure, ACM CCS DeFi workshop 2021, DOI 10.1145/3464967.3488588
• Andrew Huang, Infra-Red In-Situ inspection of silicon, arXiv 2303.07406, 2023, preprint, not peer reviewed
• Chainalysis, 18 December 2025, 2025 crypto theft reaches 3.4 billion dollars
• European Commission, Cyber Resilience Act, Regulation (EU) 2024/2847
• FBI Internet Crime Complaint Center, annual reports index, including the 2025 report and its cryptocurrency supplement