Menu
Breach Value Calculator

Encrypted USB Drive Vulnerability: Lenovo ThinkPlus and What to Check

A Lenovo ThinkPlus drive sold as AES-256 encrypted held a plain SD card. AI-aided research found fixed keys or hidden accounts in other USB products.

By CyphrQ.ai

Written with AI assistance and human editorial review. Every figure linked to its source.

Published 12 min read

Opened up, the highlighted part is an ordinary SD card: the fingerprint pad only switched the USB connection on, so the data on the card was never encrypted.

Can you trust an "encrypted" USB drive? Not on the label alone. Xusheng Li of Vector 35 found that Lenovo's ThinkPlus FU100, a fingerprint drive sold as AES-256 encrypted, stored files on an ordinary, unencrypted SD card: an encrypted USB drive vulnerability in its plainest form.

Lenovo published three CVEs on 14 January 2026, including CVE-2025-13453 for missing encryption (CWE-311). Li then used AI to find fixed keys or hidden accounts in four other products. In Netac's NetacFileLock software, the AI reached the main flaw in seven minutes.

Vector 35 is the company behind the Binary Ninja reverse-engineering tool. Lenovo says it has fixed its drives, and that it sold the ThinkPlus products only in mainland China.

The lesson travels further than one market. "Hardware encryption" on a box is a claim, not proof. Below: what was found, what Lenovo said, and what to check before you buy or keep using a drive like this.

What was the Lenovo ThinkPlus encrypted USB drive vulnerability?

Li's timeline, in his talk slides on GitHub, runs from his first report to Lenovo in August 2025, through three CVEs in January 2026, to his talk at the RE//verse conference in March 2026. Lenovo's statement to Tom's Hardware gives the same August 2025 report date, and the CVE-2025-13454 record credits Li.

His slides describe two problems in particular:

  • ThinkPlus FU100. Sold as an AES-256 encrypted fingerprint drive, it was, in the slides' words, "just an SD card in a fancy case". The fingerprint check only switched on the USB connection; the data itself was not encrypted.
  • ThinkPlus TSD303. This encrypted SSD did encrypt the data, but it sent the password to the computer in clear text.

Lenovo, acting as the CVE Numbering Authority (CNA) for its own products, published three records on 14 January 2026. They cover the FU100, FU200, TU800 and TSD303, listed as "Gen 1", and point to Lenovo's advisory LEN-210545, which rates the flaws High severity, with information disclosure as the potential impact.

CVE What the record says Weakness (CWE) CVSS v4.0 CVSS v3.1
CVE-2025-13453 Some ThinkPlus USB drives could let someone with physical access read stored data CWE-311, Missing Encryption of Sensitive Data 5.1 Medium 4.6 Medium
CVE-2025-13454 ThinkPlus configuration software could let a local authenticated user access sensitive device information CWE-319, Cleartext Transmission of Sensitive Information 6.8 Medium 5.5 Medium
CVE-2025-13455 ThinkPlus configuration software could let a local authenticated user bypass device authentication and enrol an untrusted fingerprint CWE-290, Authentication Bypass by Spoofing 7.3 High 7.8 High

The "Medium" score on the first record can mislead. It is scored for an attacker who needs the drive in hand. For a USB drive, that is exactly the case that matters: a lost or stolen stick.

What did Lenovo say?

Lenovo's statement, quoted in the Tom's Hardware update of 3 November 2026, makes these points:

Point Lenovo's position, as reported
Market The products were designed for the China market and only available from Lenovo in mainland China
Disclosure Reported to Lenovo in August 2025 and handled through coordinated disclosure
Fix Remediated through hardware and software upgrades; advisory and three CVEs published in January 2026
Customers Can contact Lenovo Support to arrange a replacement

Lenovo's advisory says it has released a hardware-level fix. Affected owners can apply for a free upgrade through a mainland China after-sales line; Lenovo checks the serial number and offers a send-back replacement or an on-site firmware upgrade. It advises backing up important data and strongly recommends deleting or formatting the drive before sending it.

The advisory does not mention the stopgap. Li's talk slides say a temporary fix for the FU100 was available by 27 August 2025, with an epoxy cover over the SD card, and a proper fix to follow in December.

How did AI find flaws in other encrypted USB tools in minutes?

On 22 October 2026, Li published a follow-up on the Binary Ninja blog. This time he worked with an AI coding agent, OpenAI's Codex, connected to Binary Ninja through MCP (the Model Context Protocol, a standard way to give an AI agent access to a tool), plus time-travel debugging for one drive. Time-travel debugging records a program's run so an analyst can step backwards and forwards through it.

None of these products is Lenovo's. One is a hardware drive; the other three are software that locks or copy-protects files on a USB stick. In each case, the user's password did not actually protect the files.

Product Vendor and type What Li found
YutianDun Hardware drive sold by SecZure Hidden built-in accounts with fixed passwords could decrypt files without the user's password
NetacFileLock Netac's file-encryption software AES-256 used a fixed key embedded in the program, so the user's password was not used for the key. The AI also found the password written to logs, ECB-style encryption and MD5
KakaSoft USB Copy Protection KakaSoft's copy-protection software In its "Highest" mode, the key came from a modified MD5 with a fixed internal value, so files could be decrypted without the password
Nexcopy Copy Secure Nexcopy's copy-protection software, with a viewer app The content key sat in a configuration file protected by a fixed certificate; the password was only a check in the viewer app

Li describes the SecZure accounts as a backdoor in the technical sense, because the fixed passwords open files without the user's password. He adds: "That does not imply malicious intent."

Where the seven minutes comes from. NetacFileLock was not new to Li. He had already broken it for his March talk, and the AI run was a rerun: "Within seven minutes, Codex reached the main flaw: the program initialized AES-256 with a fixed 32-byte key embedded in the executable." In his words, the AI "reproduced my earlier research in seven minutes using static analysis." Both lines are about NetacFileLock, not the Lenovo drives.

The AI was not working alone. Li reports that the KakaSoft and Nexcopy results took several follow-up prompts from him, and that he had to press the AI for detail.

Disclosure. Li writes that he did not follow the usual coordinated disclosure process for the bugs in his October post, arguing that a device already in an adversary's hands cannot easily be recalled or updated. Tom's Hardware reports that KakaSoft, Nexcopy and SecZure were not warned in advance. Netac had been told earlier: Li's slides show he reported NetacFileLock to Netac on 19 October 2025, got no response, and that no patch was available when he disclosed it in his March 2026 talk, after more than 90 days.

This fits a wider pattern of AI speeding up vulnerability discovery, which we covered in our post on DARPA's AI Cyber Challenge. The difference here is the target: consumer products whose security claim is the whole point of buying them.

Why can "hardware encryption" be only a marketing claim?

"AES-256" names a strong cipher. It says nothing about how the key is made, where it is kept, or whether the cipher is used at all.

In the cases above, the cipher was rarely the weak point. The failures were around it:

  • No encryption at all, behind a fingerprint reader (FU100, per Li's slides; CVE-2025-13453).
  • The password sent in the clear, where software on the computer could read it (TSD303, per Li's slides; CVE-2025-13454).
  • A fixed key shipped in every copy of the software, so the password is a gate, not a key (NetacFileLock).
  • Hidden accounts that open any drive (SecZure YutianDun).

A fingerprint sensor or keypad looks like security. If it only decides whether the computer sees the storage, anyone who opens the case can read the chip directly.

How do I know an encrypted USB drive is really secure?

  1. Look for an independent validation, not a cipher name. In the US and Canada, the Cryptographic Module Validation Program (CMVP), run jointly by NIST and the Canadian Centre for Cyber Security, validates cryptographic modules against FIPS 140-3. Ask for the certificate number and check it on the program's validated-modules list.
  2. Check the certificate covers this product. The module name and version on the certificate should match the drive you are buying, not a component inside it or an older model.
  3. Ask where the key comes from. The right answer is a key generated on the device and protected by your password or PIN. A key that is the same for every unit, or held in desktop software, is the wrong answer.
  4. Ask what happens on wrong guesses. A drive should limit attempts and wipe or lock itself after too many.
  5. Treat software-only "USB locking" tools as convenience, not protection, unless the vendor publishes how keys are handled and someone independent has checked it.
  6. Ask about fixes. Can the firmware be updated? Is there a security contact and a published advisory history?

What should I do if I own an affected drive?

  1. ThinkPlus FU100, FU200, TU800 or TSD303 (Gen 1): apply to Lenovo for the free replacement or firmware upgrade its advisory offers. Back up your files and wipe the drive first, as Lenovo recommends.
  2. SecZure YutianDun, NetacFileLock, KakaSoft or Nexcopy: assume the password does not protect your files, based on Li's findings. Move sensitive files off the drive and protect them with encryption you control.
  3. Any drive you are unsure about: keep sensitive data on it only if it is also encrypted with a tool you trust, such as the full-disk or container encryption built into your operating system.
  4. If a drive with sensitive data is already lost, treat the data as exposed and follow your incident or data-protection process.

Going deeper: for security and IT teams

Standards and identifiers. The three Lenovo records map to distinct weaknesses: CVE-2025-13453 is CWE-311 with vector CVSS:4.0/AV:P/AC:L/AT:N/PR:N/UI:N/VC:H/VI:N/VA:N/SC:N/SI:N/SA:N; CVE-2025-13454 is CWE-319, AV:L/AC:L/AT:N/PR:L; CVE-2025-13455 is CWE-290, AV:L/AC:L/AT:P/PR:L with high confidentiality, integrity and availability impact. The CVE-2025-13454 record credits Xusheng Li of Vector 35. For procurement, the relevant standard is FIPS 140-3, validated through CMVP, which began accepting FIPS 140-3 submissions on 22 September 2020, according to NIST's FIPS 140-3 transition page.

How it fails. A sound encrypted drive has a media encryption key generated on the device, never exported, and wrapped by a key derived from the user's secret. Authentication then unwraps the key; without the secret, the stored ciphertext is useless. Every failure in this story breaks that chain at a different link. The FU100 had no media encryption. The TSD303 encrypted, but put the verifying secret on the USB bus, where host software can observe it. NetacFileLock and KakaSoft derived the key from a constant, so the password only gated a user interface. Nexcopy stored the content key under a fixed certificate in the viewer. SecZure added accounts that bypass the user's secret by design. ECB-style block encryption, also reported for NetacFileLock, leaks patterns because identical plaintext blocks produce identical ciphertext.

Threat model. The realistic attacker holds the device: a finder, a thief or an insider. CVSS physical-access scoring understates this for removable media. Li's point that a device in an adversary's hands cannot easily be recalled or updated cuts both ways: firmware fixes do not reach drives already lost, so data written before a fix should be treated as unprotected. For CVE-2025-13454 and -13455 the attacker needs local, authenticated access to a host the drive is plugged into, which puts malware on a shared or compromised workstation in scope.

In practice: the speed of discovery has changed. Li reports that Codex "reached the main flaw" in NetacFileLock within seven minutes, using static analysis through Binary Ninja's MCP connection (Li's Binary Ninja post). He had first broken the same program by hand for his March 2026 talk on time-travel debugging (talk slides). Assume any key or credential embedded in a client binary will be found quickly. "Nobody will reverse-engineer it" is no longer a defensible design assumption, if it ever was.

In practice: severity is not the same as relevance. CVE-2025-13453 scores 5.1 under CVSS v4.0, yet it means the FU100's stored data was readable by anyone with the drive. Rank removable-media findings by the asset class they protect, not by the base score alone.

Controls.

  • Keep an approved-device list for removable storage and enforce it through your endpoint management tool's device-control policy, allowing only models with a current CMVP certificate.
  • Where an approved drive is not available, require operating-system encryption on removable media before data is written.
  • Remove "USB lock" utilities from the approved software list unless the vendor documents key handling and offers an independent assessment.
  • Write lost-device handling into your incident process: who reports, what is assumed exposed, and when notification duties start.

Detection and evidence. Inventory removable storage by vendor and product ID through endpoint logs and look for the ThinkPlus Gen 1 models and the software products above. Watch for the configuration utilities in software inventory. Test approved drives yourself: confirm the drive refuses access without its secret, check that repeated wrong attempts trigger lockout, and ask the vendor for its CMVP certificate and security policy document.

What to ask vendors. Is the media encryption key unique per device and generated on it? Is the cryptographic module FIPS 140-3 validated, and which certificate number? Does any account, service mode or recovery path decrypt data without the user's secret? How are firmware fixes delivered, and how did you handle your last reported vulnerability?

Key takeaways

  • Lenovo's ThinkPlus FU100, sold as an AES-256 encrypted fingerprint drive, held an unencrypted SD card, according to Li's slides and CVE-2025-13453.
  • Lenovo published CVE-2025-13453, -13454 and -13455 on 14 January 2026 for four Gen 1 ThinkPlus models, which it says it sold only in mainland China, and offers free replacements or firmware upgrades.
  • In a follow-up dated 22 October 2026, Li reports that AI helped him find fixed keys or hidden accounts in four other products. In NetacFileLock it reached the main flaw in seven minutes.
  • Netac had no patch when Li disclosed its flaw in March 2026, more than 90 days after his report.
  • He calls the SecZure accounts a backdoor in the technical sense and says that does not imply malicious intent.
  • Buy on independent validation (a FIPS 140-3 certificate that names the product), not on a cipher name on the box.

Sources

Read on 4 November 2026.

Spotted a mistake? Report a correction.