Ledger CTO Charles Guillemet: Reproducing a Patched Vulnerability is Not an Intrusion
Ledger CTO Charles Guillemet responded that reproducing a patched vulnerability does not constitute "hacking Ledger." He referred to a thread describing a flaw in an old version of the Ethereum application, which was discovered by the company's security processes and patched in Ethereum application version 1.22.2 released on August 13, prior to this post.
He clarified: No users were hacked, and there was no exploitation in the wild; running exploits on old versions after the patch was issued is merely a laboratory exercise, not a new discovery. The conclusion remains unchanged: all software has flaws, and the key is what to do after they are discovered. The model involves the internal security team Ledger Donjon and external researchers identifying flaws, fixing every device in the field, and issuing public announcements.
He urged users to upgrade their Ethereum applications to version 1.22.3 or higher via Ledger Wallet and to verify the application version on their devices, not just the firmware. Applications and firmware are updated separately. The security announcement later specified that version 1.22.2 added two locks, rejecting new signing sessions during review, and invalidating confirmations if the screen display does not match the memory state. On August 21, Secure SDK 26.6.1 further repaired the underlying issues, and applications were rebuilt according to the corrected software.
The public controversy stems from two lines. Security company TestMachine disclosed a transaction replacement risk in the clear signing process between August 21 and 24, stating that malicious decentralized applications could swap out pending signing data while users are focused on the screen; Ledger claimed that Donjon discovered it first using internal tools, and the other party only contacted the bounty program after the patch went live, not following the complete responsible disclosure process. OneKey founder Yishi Wang later stated that their security team reproduced a race condition between display logic and transaction buffer in the lab for version 1.22.1.
Ledger Donjon also issued announcement No. 024: the generic transaction parser wrote a controllable 16-bit array element count into an 8-bit countdown field, truncating 257 elements to 1, with the device only verifying the last item while signing the entire transaction; this path existed from versions 1.19.0 to 1.22.2, with the fix in version 1.22.3. Announcement No. 025 covered the exchange path: it verified the address and amount but did not verify the nature of the operation, allowing token authorizations to be mistaken for payment sign-offs. The company stated that there was no evidence of exploitation in the wild for both cases. Affected models reported include Flex, Nano X, Nano S Plus, Stax, and Apex; the Nano S was indicated as not applicable to this Ethereum flaw.
Trust is being repriced. Buyers need to confirm that "the screen is what is signed" for holders, while sellers are the patch channels and competing product disclosures. Funds have not flowed out due to theft, and attention has shifted from device sales to application version numbers. Beneficiaries are users who have upgraded and manufacturers who control update distribution; those under pressure are holders who only upgrade firmware without upgrading on-chain applications, and those who frame laboratory reproductions as intrusions. This is a dispute over the timing of disclosures, not a large-scale incident on-chain.
Source: Public Information
ABAB AI Insight
Ledger builds its security on "chip isolation and screen confirmation," facing repeated questions since 2018 about recovery phrase leaks and actual losses from the 2023 Connect Kit supply chain poisoning. Donjon is its laboratory for internalizing vulnerabilities: first discovering them internally, then silently upgrading applications, and finally issuing announcements. This time, the 1.22.2 update log only mentioned "security issues," with the announcement coming about two weeks after the code tag, effectively pulling the disclosure rhythm back to the manufacturer’s calendar rather than the researcher’s calendar.
The capital path is hardware premiums exchanged for updatable software. Money continues to flow into devices, subscriptions, and bounties, with resources diverted to AI-assisted vulnerability discovery and SDK reconstruction. The motivation is to maintain the product boundary of "not updated does not count as compromised": private keys remain in the secure element, while the challenge is at the level of clear signing. OneKey’s laboratory reproduction seizes the narrative, as the wallet market replaces advertising space with security posts, similar to antivirus vendors attracting new customers through virus demonstrations.
Similar to the 2010s browser vendors competing with vulnerability brokers over whether a fixed issue counts as a vulnerability, it is also akin to the Trezor and Coldcard camps attacking closed-source upgrade models with open-source firmware. The industry is in a control phase: hardware wallets have shifted from "never connected vaults" to "must be regularly updated signers." Not updating applications equates to voluntarily remaining on an old attack surface, with manufacturers pushing responsibility back onto users.
The essence is a transfer of pricing power. The screen once priced transaction content, while race conditions and truncated counts allow "what is seen" and "what is signed" to be separated. Whoever controls application distribution can rewrite vulnerabilities from product accidents to user upgrades not being performed. When fixes can be silently pushed to all devices in the field, the discovery power shifts from external researchers to internal pipelines, leaving public posts with only attention, no longer automatically equating to new risks.
ABAB News · Cognitive Laws
Patches have been issued, reproducing is just a drill
New firmware does not equal new applications
Content on the screen does not have to equal what is signed