Ledger has patched a vulnerability in its Ethereum hardware-wallet application that could have allowed a malicious dApp to swap a transaction during the approval process while the device continued displaying the transaction the user had originally reviewed.

That on-screen review is the foundation of a hardware wallet's security model: private keys stay isolated on the device, and the transaction shown on its screen is supposed to be exactly the one signed, even when the connected computer or browser is compromised.

Security firm TestMachine disclosed the signature-substitution flaw on August 21 after reproducing it on Ledger hardware (X post). Ledger had already shipped Ethereum app version 1.22.2 on August 12, with the release history listing security fixes for the update (changelog).

Malicious dApp Could Alter the Transaction Before Signing

The flaw affected the command flow between a connected dApp and Ledger's Ethereum application. A malicious application with direct device access through WebHID could interfere while a transaction review remained open, replacing the signing context before the user approved it. WebHID is a browser interface that lets web pages communicate with hardware devices directly, which is how a web-based application could reach the wallet.

The device could therefore keep displaying the original transaction while preparing a different transaction for the final signature. TestMachine demonstrated the issue using a benign ETH transfer that could be substituted with a token approval — a transaction type that can grant another address or contract permission to spend the user's tokens — creating a path for users to authorize something other than the transaction shown on the hardware screen.

Version 1.22.2 blocks competing signing commands while a review is pending and tightens how the application maintains signing state during the approval flow. The fix preserves the link between the transaction displayed on the secure screen and the transaction ultimately signed by the device.

Ledger Says Donjon Found the Bug Before Disclosure

Ledger CTO Charles Guillemet said the company's Donjon security team had already identified and fixed the bug roughly two weeks before TestMachine published its findings, using AI-assisted security tools during the review (X post).

Guillemet also criticized the disclosure process, saying TestMachine contacted Ledger's bounty program only after the patch had already shipped and then published its findings without completing the normal responsible-disclosure process. Under coordinated disclosure, researchers report flaws privately and give vendors time to ship a fix before details go public, and that sequencing is at the heart of the dispute.

Users running current device firmware and Ethereum app version 1.22.2 are protected against the flaw.

The discovery adds another example of AI entering wallet-security review. A recent Bitcoin Red Team audit used AI-assisted analysis alongside human researchers to identify thousands of potential security issues across Bitcoin wallets and infrastructure.

Hardware Wallet Vendors Push Security Updates

Ledger's fix arrives less than a week after BitBox released firmware 9.26.5 to patch two severe vulnerabilities, including a memory-corruption flaw that could allow arbitrary code execution under specific conditions.

Hardware-wallet security has faced heavier scrutiny since the Coldcard seed-generation failure exposed wallets created with insufficient randomness. Because a wallet's seed phrase derives its private keys, weak randomness can leave those keys reproducible by others. The resulting Bitcoin drains were estimated at as much as $132 million as multiple attackers raced to identify vulnerable addresses.

Ledger's Ethereum app 1.22.2, released August 12, is the patched version covering the newly disclosed signing flaw; earlier app versions do not include the fix.