Attack ChainAIDerived
Hypothetical attack flow derived from CVE metadata – not a confirmed exploit chain
-
AccessCraft malicious dsh.bundle.patch value
-
ExploitInduce user to load bundle patch
-
ExecutionPath traversal escapes bundle dir in loadProfile
-
ImpactRead out-of-bundle file contents
Vulnerability Assessment
| Weakness |
CWE
OWASP Top 10:2025
A01:2025 - Broken Access Control
ASVS 5.0
V5 - File Handling Path canonicalization, traversal prevention
MITRE ATT&CK
2 techniques across 2 tactics: Discovery, Collection The technique list with ATT&CK links and the kill-chain view is part of the Basic plan and up. |
| Exploitation Conditions | Exploitation requires the attacker to supply a crafted dsh.bundle.patch argument that is processed by the loadProfile function of the Bundle Patch Handler (packages/boot/app-boot/src/profile.ts). Per the CVSS 4.0 vector, no privileges are required (PR:N) but passive user interaction is required (UI:P) - a user must load or apply the malicious bundle patch - and the attack has high complexity (AC:H), with VulDB explicitly describing exploitability as 'difficult.' Impact is limited to low confidentiality (VC:L: reading files outside the intended bundle path); there is no integrity or availability impact. Limiting factors: the product is a niche release-candidate AI harness with minimal deployment, the high-complexity barrier, and the requirement that a victim load attacker-controlled patch input. |
| Impact | This is a low-real-risk issue despite public exploit availability. The vendor CVSS 4.0 base score is 2.3 (Low), and the vector (AV:N/AC:H/AT:N/PR:N/UI:P/VC:L/VI:N/VA:N) is internally consistent with the description: network-reachable but high attack complexity ('the exploitability is described as difficult'), requiring passive user interaction, and yielding only low confidentiality impact (out-of-bundle file read) with no integrity or availability impact. VulDB is the sole source; there is no EPSS score and no CISA KEV listing, so there is no evidence of automated or widespread exploitation. A public exploit exists (VulDB-referenced GitHub discussion #3354), but POC availability is a distinct signal from active exploitation and does not by itself elevate priority given the AC:H barrier and minimal impact. The affected product - DeepSeek deepseek-harness, an early-stage (up to 0.1.7-rc.2 release-candidate) TS/Node AI agent harness - has a tiny deployment footprint, further limiting real-world exposure. Net: verifiable but low priority; not a candidate for emergency patching. |
| Exploit ScenarioDerived | A remote attacker crafts a malicious dsh.bundle.patch value containing directory-traversal sequences and induces a user of the deepseek-harness to load it, causing loadProfile in profile.ts to resolve a path outside the intended bundle directory and disclose file contents. A public proof-of-concept exists (VulDB-referenced GitHub discussion #3354). Exploitation is rated difficult (high attack complexity) and depends on user interaction, so opportunistic mass exploitation is unlikely. |
Get an e-mail when this CVE is exploited, patched or re-scored. Free account.
Alert meAnalysisAIv1 – Sep 28, 2026
Path traversal in the DeepSeek deepseek-harness boot tooling (through 0.1.7-rc.2) lets an attacker who can influence the dsh.bundle.patch argument read files outside the intended bundle path via the loadProfile function in packages/boot/app-boot/src/profile.ts. The issue is remotely triggerable with no privileges required (PR:N), but CVSS 4.0 rates it AC:H with UI:P, meaning a victim must load or apply the attacker-controlled bundle patch input, and VulDB itself characterizes exploitability as 'difficult'. Publicly available exploit code exists, but impact is confined to low confidentiality (CVSS 4.0 score 2.3; independent assessment CVSS 3.1 2.0/2.3-class finding AV:N/AC:H/PR:N/UI:R/S:U/C:L/I:N/A:N), so this is a low-real-risk issue despite the public proof-of-concept.
RemediationAI
No vendor-released patch was identified at time of analysis - the vendor was contacted before disclosure and did not respond, so treat 0.1.7-rc.2 and earlier as unfixed and avoid deploying this release-candidate harness in environments where it processes untrusted bundles. Until a fixed release appears, the most effective compensating control is to never load or apply bundle patches from untrusted or externally sourced origins, since exploitation depends on the victim supplying the malicious dsh.bundle.patch input (UI:P); treat any bundle patch referencing dsh.bundle.patch as a code/config artifact requiring review. Where the harness must be used, run it with a least-privilege service account and a restricted filesystem view (container or chroot with only the required bundle directories mounted), which blocks traversal reads outside the bundle path at the cost of added operational overhead. Optionally, sanitize or whitelist the dsh.bundle.patch value in the configuration (reject absolute paths and any '..' segments) before boot; this is a local customization that must be re-applied after each upgrade. Monitoring the DeepSeek repository discussions/3354 and VulDB entry 410989 for a patched release is recommended, and organizations unable to apply any of these controls should disable the bundle patch loading path or remove the harness from the affected host.
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
The flaw is a classic CWE-22 path traversal in the Bundle Patch Handler of the deepseek-harness boot layer. The loadProfile routine in packages/boot/app-boot/src/profile.ts consumes a configuration value named dsh.bundle.patch and uses it to locate/apply a bundle patch file, without adequately normalizing or constraining the resulting path (e.g., traversal sequences such as ../ escaping the intended bundle directory). Because the value is attacker-controllable through the bundle patch input and the component is invoked during boot profile loading, a crafted value can direct reads to arbitrary files accessible to the harness process. The CPE cpe:2.3:a:deepseek:deepseek-harness:*:*:*:*:*:*:*:* matches all published versions up to and including 0.1.7-rc.2, a release-candidate line of an AI harness. Per the CVSS 4.0 vector (AV:N/AC:H/AT:N/PR:N/UI:P/VC:L/VI:N/VA:N/SC:N/SI:N/SA:N/E:P), only confidentiality with low impact is at stake - no integrity or availability effect - and the E:P metric confirms a public proof-of-concept rather than demonstrated real-world exploitation.
Know when Deepseek Harness 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:N/AC:H/AT:N/PR:N/UI:P/VC:L/VI:N/VA:N/SC:N/SI:N/SA:N/E:P
Network-reachable but 'difficult' exploitation (AC:H) and requires a user to load the malicious patch (UI:R); path traversal yields only file-read disclosure (C:L), no I/A impact.
Primary rating from Vendor (VulDB).
CVSS VectorVendor: VulDB
Lifecycle Timeline
2DescriptionCVE.org
A security flaw has been discovered in DeepSeek deepseek-harness up to 0.1.7-rc.2. The affected element is the function loadProfile of the file packages/boot/app-boot/src/profile.ts of the component Bundle Patch Handler. The manipulation of the argument dsh.bundle.patch results in path traversal. The attack may be performed from remote. The attack requires a high level of complexity. The exploitability is described as difficult. The exploit has been released to the public and may be used for attacks. The vendor was contacted early about this disclosure but did not respond in any way.
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
Path traversal exploits occur when applications use user-controlled input to construct file system paths without proper validation.
More in Deepseek Harness
View allDeepSeek Harness versions up to and including 0.1.1-rc.2 do not properly enforce the intended behavioral workflow within
Local sandbox escape in deepseek-ai deepseek-harness up to 0.1.7-rc.2 lets a low-privileged attacker who already has cod
Local information disclosure in deepseek-ai DeepSeek Harness through 0.1.5-rc.3 allows an authenticated local attacker w
EUVD-2026-88438
GHSA-rh7q-h525-wx9f
NVD
CVE.org
EPSS
Vulnrichment
VulDB
Exploit-DB