Vulnerabilities

Ray prior to 2.56.0 contains an unsafe deserialization vulnerability in the WebDataset reader that allows attackers to achieve remote code execution by supplying a malicious tar archive to the read_webdataset() function. The _default_decoder() function in webdataset_datasource.py unconditionally calls pickle.loads() on tar entries with .pkl/.pickle extensions and torch.load() with weights_only=False on .pt/.pth entries, executing arbitrary code inside Ray remote workers on every worker that processes the malicious archive.

Ray is an AI compute engine. From version 2.54.0 to before version 2.55.0, Ray Data registers custom Arrow extension types (ray.data.arrow_tensor, ray.data.arrow_tensor_v2, ray.data.arrow_variable_shaped_tensor) globally in PyArrow. When PyArrow reads a Parquet file containing one of these extension types, it calls __arrow_ext_deserialize__ on the field's metadata bytes. Ray's implementation passes these bytes directly to cloudpickle.loads(), achieving arbitrary code execution during schema parsing, before any row data is read. This issue has been patched in version 2.55.0.

A path traversal vulnerability was identified in Ray Dashboard (default port 8265) in Ray versions prior to 2.8.1. Due to improper validation and sanitization of user-supplied paths in the static file handling mechanism, an attacker can use traversal sequences (e.g., ../) to access files outside the intended static directory, resulting in local file disclosure.

Ray is an AI compute engine. In versions 2.53.0 and below, thedashboard HTTP server blocks browser-origin POST/PUT but does not cover DELETE, and key DELETE endpoints are unauthenticated by default. If the dashboard/agent is reachable (e.g., --dashboard-host=0.0.0.0), a web page via DNS rebinding or same-network access can issue DELETE requests that shut down Serve or delete jobs without user interaction. This is a drive-by availability impact. The fix for this vulnerability is to update to Ray 2.54.0 or higher.

Ray, the open-source AI compute engine, is vulnerable to a critical remote code execution flaw (CVE-2025-62593, CWE-94/CWE-352) in versions prior to 2.52.0, caused by an insufficient guard against browser-based attacks: the software distinguishes browser traffic only by checking that the User-Agent header starts with 'Mozilla', but the fetch specification allows that header to be modified. An attacker can combine DNS rebinding with a crafted User-Agent so that a developer's Firefox or Safari browser silently sends malicious requests to locally running Ray services after the developer visits an attacker-controlled website or is served a malicious advertisement (malvertising). Successful exploitation yields full remote code execution on the machine running Ray, with high confidentiality, integrity and availability impact reflected in the CVSS 4.0 score of 9.4. Affected users are developers running Ray as a development tool on any version before 2.52.0, which is the fixed release. The flaw is being actively exploited in the wild: CISA added it to the Known Exploited Vulnerabilities catalog on 2026-08-17, a public proof of concept is available in the project's GitHub security advisory (GHSA-q279-jhrf-cc6v), and EPSS estimates a 16.9% probability of exploitation in the next 30 days.

Do: Upgrade Ray to version 2.52.0 or later immediately, in line with CISA KEV and BOD 26-04 patching requirements. Until patched, avoid browsing untrusted websites (including ad-serving pages) while Ray development services are running, and restrict network access to locally running Ray services. Focus triage on machines where developers use Firefox or Safari alongside Ray, and hunt for signs of compromise since the flaw is actively exploited.

Anyscale Ray 2.6.3 and 2.8.0 allows a remote attacker to execute arbitrary code via the job submission API. NOTE: the vendor's position is that this report is irrelevant because Ray, as stated in its documentation, is not intended for use outside of a strictly controlled network environment. (Also, within that environment, customers at version 2.52.0 and later can choose to use token authentication.)