Attack ChainAIDerived
Hypothetical attack flow derived from CVE metadata – not a confirmed exploit chain
-
AccessObtain leaked providers export JSON file
-
ExploitRead hardcoded password 'lobsterai-APP' from public source
-
ExecutionDerive AES-GCM key via PBKDF2 with embedded salt
-
ImpactDecrypt and recover all provider API keys
Vulnerability Assessment
| Weakness |
CWE
OWASP Top 10:2025
A04:2025 - Cryptographic Failures
ASVS 5.0
V11 - Cryptography Key management, no hard-coded keys
|
| Exploitation Conditions | Exploitation requires the attacker to obtain a providers export file produced by LobsterAI's 'export providers' feature (named like 'lobsterai-providers-<date>.json', version 2, encryption keySource 'password'). The file encrypts each provider's apiKey with the compile-time constant EXPORT_PASSWORD ('lobsterai-APP') rather than a user-chosen secret, so no password guessing or brute force is needed. Limiting factors that reduce real-world exposure: a user must have actively used the export function (UI:P), the attacker must gain access to the resulting file through a separate channel (leak, shared storage, repository, or backup), and the vulnerability is not reachable over the network against the running application (AV:L). Versions before 2026.10.16 are affected; 2026.10.16 and later remediate it. |
| Impact | This is a genuine but moderate-priority confidentiality issue, not a high-urgency remote threat. The CVSS 4.0 vector (AV:L/AC:L/AT:N/PR:N/UI:P/VC:H) scores 6.8 and the signals are internally consistent: the local attack vector and passive user-interaction requirement correctly reflect that an attacker cannot reach the flaw remotely - they must first obtain a providers export file that a user chose to generate. The high confidentiality impact (VC:H) with no integrity or availability impact is appropriate, since the only consequence is disclosure of third-party AI provider API keys. No EPSS score, CISA KEV listing, or published proof-of-concept exploit was provided in the input, so there is no public exploit identified at time of analysis and no evidence of active exploitation; however, the hardcoded constant 'lobsterai-APP' and the exact source locations are publicly visible on GitHub, meaning the decryption technique is trivially reproducible by anyone who reads the code. The real risk is therefore driven by data exposure (leaked, shared, or backed-up export files) rather than by direct attack against the running application, and it is bounded by the relatively small deployment footprint of this product. Treat as a routine patch-when-convenient item that becomes urgent only if export files are known to have circulated. |
| Exploit ScenarioDerived | A LobsterAI user exports their provider configuration and the resulting JSON file is leaked - committed to a repository, attached to a support ticket, or synced to cloud backup. An attacker obtains the file and, using the publicly known hardcoded password 'lobsterai-APP', derives the AES-GCM key via PBKDF2 from the embedded salt and decrypts every stored third-party AI provider API key. No proof-of-concept is published, but the technique is fully described by the vulnerable source and requires only the leaked file plus local offline computation. |
Get an e-mail when this CVE is exploited, patched or re-scored. Free account.
Alert meAnalysisAIv1 – Oct 18, 2026
LobsterAI before 2026.10.16 encrypts exported provider API keys with a hardcoded compile-time constant ('lobsterai-APP') instead of a user-supplied secret, so anyone who obtains a provider export file can derive the AES-GCM key via PBKDF2 with the embedded salt and recover every third-party AI provider API key in cleartext. The CVSS 4.0 vector (AV:L/PR:N/UI:P, VC:H only) reflects that exploitation is unauthenticated in principle but requires the victim to have used the export feature and the attacker to separately obtain the resulting JSON file - this is a moderate-priority confidentiality issue, not a high-urgency remote threat, and no public exploit code was identified at time of analysis. Netease Youdao has released a fix in LobsterAI 2026.10.16.
RemediationAI
Vendor-released patch: upgrade to LobsterAI 2026.10.16 or later, which replaces the hardcoded-constant export path with the ProviderTransferDialog/useProviderTransfer implementation (commit 1eb28c40d7e42d6cbade07874e6f6c068b6a9a64); see https://github.com/netease-youdao/LobsterAI/releases/tag/2026.10.16. Upgrading alone does not undo the exposure, so also rotate or revoke every third-party AI provider API key that was present in any previously exported providers JSON, and delete or quarantine those files from shared storage, repositories, ticketing systems, chat history, and backups (treat them as plaintext secrets - the AES-GCM layer offers no protection). Until upgrading, stop using the 'export providers' feature and block it via endpoint policy if it must be prevented, accepting that this breaks legitimate migration workflows. Searching code repositories and file shares for the filename pattern 'lobsterai-providers-*.json' and for export payloads containing "keySource": "password" will identify leaked exports; removing such a file from version control history requires a history rewrite, and any key in it must be considered compromised from the moment it was committed.
Affected VersionsENISA
Affected Packages
Source: OSV (Open Source Vulnerabilities)
Affects your stack
Affects 1 package in euvd
Checking this CVE against your own stacks and servers is part of the Basic plan and up.
Technical ContextAI · show details
CWE-321 (Use of Hard-coded Cryptographic Key) is realized here in the renderer layer of the LobsterAI Electron desktop application. The settings UI imported an EXPORT_PASSWORD constant from src/renderer/constants/app.ts (line 4) and used it as DEFAULT_EXPORT_PASSWORD when building a version-2 providers export payload; each provider's apiKey was wrapped with encryptWithPassword(), producing an AES-GCM ciphertext whose key is derived via PBKDF2 from that constant plus a salt stored alongside the data in the export file. Because both the passphrase and the derivation parameters travel in (or are fixed for) the file, the encryption provides no confidentiality against anyone holding the export - no password guessing or brute force is required, which is why the assessed vector is AV:L/AC:L/PR:N with UI:R and only a confidentiality (VC:H) impact. The patch commit 1eb28c40d7e42d6cbade07874e6f6c068b6a9a64 deletes the hardcoded export/import logic and the ProviderExportEntry/ProvidersExportPayload types from Settings.tsx and moves the functionality into a ProviderTransferDialog component and a useProviderTransfer hook, changing the key source away from a static constant. CPE cpe:2.3:a:netease-youdao:lobsterai:* confirms all builds prior to the 2026.10.16 release tag are affected.
Know when Lobsterai gets its next vulnerability
- Watch products and get an email when something new is published
- Check this CVE against your own stack and servers
- Exploit and PoC references, threat intelligence from 22 sources
No credit card · 7-day full trial
Severity by source
CVSS:4.0/AV:L/AC:L/AT:N/PR:N/UI:P/VC:H/VI:N/VA:N/SC:N/SI:N/SA:N
Attacker needs the locally obtained export file (AV:L) created by a user action (UI:R); decryption with the public constant needs no privileges (PR:N), and only API keys are disclosed (C:H, I:N, A:N).
Primary rating from Vendor (VulnCheck).
CVSS VectorVendor: VulnCheck
DescriptionCVE.org
LobsterAI before 2026.10.16 contains a hardcoded password vulnerability that allows attackers to decrypt exported provider API keys encrypted with the constant 'lobsterai-APP'. Attackers who obtain a providers export file can derive the AES-GCM key via PBKDF2 using the embedded salt and recover every third-party AI provider API key.
Threat Intelligence Overview
How this decision was reached
The engine reads each signal below and applies a fixed list of rules from top to bottom. The first rule that matches sets the recommended action – like a statement, every line is shown.
What every action level means
SSVC AssessmentENISA
Priority Score
Attack Technique
Information disclosure occurs when an application unintentionally exposes sensitive data that aids attackers in reconnaissance or directly compromises security.
Recursive deletion of arbitrary user-writable directories is possible in the LobsterAI desktop application (NetEase Youd
Server-side request forgery in netease-youdao LobsterAI affects the `buildBrowserConfig` function within the Browser Net
Information disclosure in netease-youdao LobsterAI 2026.6.10 exposes sensitive data through manipulation of the MEDIA Pa
EUVD-2026-103467
GHSA-rpw3-ccp2-4v8w
NVD
CVE.org
EPSS
Exploit-DB