Attack ChainAIDerived
Hypothetical attack flow derived from CVE metadata – not a confirmed exploit chain
-
AccessCraft checkpoint with malicious weight_map
-
ExploitVictim loads untrusted checkpoint
-
ExecutionPath traversal in load_pretrained_gr00t
-
ImpactAccess files outside checkpoint directory
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 victim to load an attacker-influenced pretrained checkpoint through the gr00tsonic Checkpoint Loader, specifically where the weight_map argument of load_pretrained_gr00t is attacker-controlled; the traversal is triggered during that checkpoint-loading operation. Per the CVSS vector no authentication is needed (PR:N), but passive user interaction is required (UI:P) - a user or automated pipeline must actually load the untrusted checkpoint, which is the primary limiting factor. Risk is further bounded to deployments running Dexbotic 0.1 or 0.2.0 that ingest checkpoints from untrusted sources; environments that only load self-produced checkpoints are not realistically exposed. No special network service or non-default feature flag is described beyond use of the checkpoint loader itself. |
| Impact | All available signals point to a genuinely low-priority issue rather than a high-CVSS-but-overrated one. The supplied CVSS 4.0 base score is only 2.1, and its metrics are internally consistent with a minor flaw: network vector (AV:N) and low complexity (AC:L) with no privileges (PR:N) but passive user interaction required (UI:P), and only Low impact to confidentiality, integrity and availability (VC:L/VI:L/VA:L) with no subsequent-system scope change (SC:N/SI:N/SA:N). There is no EPSS score in the input and no CISA KEV listing, so there is no evidence of widespread or active exploitation. The vuln.today decision label is 'monitor', the weakest urgency tier, which agrees with the low score. A public PoC reference (a poc.zip attachment) exists but is explicitly not verified as working exploit code, so it should not be treated as confirmed weaponization. The main tension is between the description's phrase 'Remote exploitation of the attack is possible' and the UI:P metric: exploitation is remote in the sense that an attacker-supplied checkpoint/weight_map can be delivered over the network, but the victim must load that untrusted checkpoint, which materially limits real-world risk. Reporting is solely third-party (VulDB); there is no vendor advisory, the project was notified but has not responded, and no fixed version is listed - verify affected scope and any fix directly with the Dexbotic project. |
| Exploit ScenarioDerived | An attacker publishes or supplies a malicious model checkpoint whose weight_map argument contains path-traversal sequences; when a developer or operator loads it through load_pretrained_gr00t in Dexbotic's gr00tsonic Checkpoint Loader, the traversal causes file access outside the intended checkpoint directory. A public PoC reference (poc.zip) is attached but is not verified as working exploit code. Exploitation hinges on the victim choosing to load the untrusted checkpoint (UI:P), so it is opportunistic rather than fully automatic. |
Get an e-mail when this CVE is exploited, patched or re-scored. Free account.
Alert meAnalysisAIv1 – Oct 28, 2026
Path traversal in the Dexbotic gr00tsonic Checkpoint Loader (dexbotic/model/gr00tsonic/gr00tsonic_arch.py, function load_pretrained_gr00t) allows an attacker who controls the weight_map argument to cause file paths outside the intended checkpoint location to be used during checkpoint loading. Dexbotic 0.1 and 0.2.0 are listed as affected (CPE cpe:2.3:a:dexmal:dexbotic:*), and exploitation is remote but requires the victim to actually load an attacker-influenced pretrained checkpoint - per the assessed CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:L/I:L/A:L vector no authentication is needed, while passive user interaction is the primary limiting factor. There is no CISA KEV entry and no confirmed active exploitation; a public PoC reference exists but it is not verified as working exploit code, and the project has been notified through an issue report without a response, so the issue is assessed as low priority (monitor).
RemediationAI
No vendor-released patch identified at time of analysis: the issue was reported to the project via https://github.com/dexmal/dexbotic/issues/103 and, per the available data, has not been answered, so no fixed version exists to cite - track that issue and https://vuldb.com/vuln/427076 for a fix. Until then, the effective control is at the loader boundary: do not load pretrained checkpoints from untrusted or third-party sources into the gr00tsonic loader, and where third-party checkpoints are unavoidable, validate every weight_map entry before load_pretrained_gr00t consumes it - reject absolute paths and any path containing '..' segments, and resolve each entry with a canonicalized prefix check confining it to the expected checkpoint directory (trade-off: sharded checkpoints that legitimately reference external paths will fail to load until their index files are rewritten). Running checkpoint loading in a sandbox or container with read-only mounts limited to the checkpoint directory contains the impact of a successful traversal, at the cost of extra operational overhead. The available data does not describe any other vendor-provided mitigation, and no patch status beyond the unanswered issue report is confirmed.
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
Dexbotic is a Python robotics/embodied-AI framework that loads pretrained model checkpoints; the affected component is the gr00tsonic checkpoint loader, specifically load_pretrained_gr00t in dexbotic/model/gr00tsonic/gr00tsonic_arch.py. The weight_map argument, which in sharded checkpoint workflows associates parameter names with checkpoint file paths, is insufficiently validated, so manipulating it yields a path traversal (CWE-22): entries supplied by the checkpoint or its index can resolve outside the intended checkpoint directory. The underlying technology is Python checkpoint deserialization and weight-shard resolution, which typically trusts the file paths recorded in checkpoint metadata. The trigger is described in the available sources only as manipulation of the weight_map argument during checkpoint loading; further mechanics of the flaw are not described in the supplied data. Impact per the assessed vector is confined to low confidentiality, integrity and availability effects on the host loading the checkpoint (S:U), not a network service compromise.
Know when Dexbotic 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:L/AT:N/PR:N/UI:P/VC:L/VI:L/VA:L/SC:N/SI:N/SA:N/E:P
An untrusted checkpoint is delivered over the network (AV:N) but must be loaded by the victim (UI:R) with no auth (PR:N); path traversal yields limited file read/write (C:L/I:L/A:L) without scope change.
Primary rating from Vendor (VulDB).
CVSS VectorVendor: VulDB
Lifecycle Timeline
3DescriptionCVE.org
A flaw has been found in Dexmal Dexbotic up to 0.2.0. Affected by this vulnerability is the function load_pretrained_gr00t of the file dexbotic/model/gr00tsonic/gr00tsonic_arch.py of the component gr00tsonic Checkpoint Loader. This manipulation of the argument weight_map causes path traversal. Remote exploitation of the attack is possible. The exploit has been published and may be used. The project was informed of the problem early through an issue report but has not responded yet.
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.
EUVD-2026-111084
GHSA-wqfq-5j5x-fgj7
NVD
CVE.org
EPSS
VulDB
Exploit-DB