CVE-2026-0599
HighSummary
Podatność w wersji 3.3.6 biblioteki huggingface/text-generation-inference pozwala nieautoryzowanym atakującym zdalnie wykorzystać nieograniczone pobieranie zewnętrznych obrazów podczas walidacji wejścia w trybie VLM. Problem występuje, gdy router skanuje wejścia w poszukiwaniu linków do obrazów w formacie Markdown i wykonuje blokujące żądanie HTTP GET, co może prowadzić do wyczerpania zasobów.
Risk Assessment
Dzięki tej podatności może dojść do nasycenia pasma sieciowego, wzrostu zużycia pamięci oraz nadmiernego wykorzystania CPU, co może doprowadzić do awarii maszyny hosta. Domyślna konfiguracja wdrożenia, która nie ma ograniczeń dotyczących użycia pamięci i autoryzacji, pogarsza sytuację.
Recommendation
Zaleca się aktualizację do wersji 3.3.7, aby usunąć tę podatność. Dodatkowo warto wprowadzić ograniczenia dotyczące użycia pamięci oraz mechanizmy autoryzacji w konfiguracji wdrożenia.
Related vulnerabilities
- CVE-2026-92229Critical
The Forminator Forms plugin for WordPress (versions up to and including 1.57.2) is vulnerable to arbitrary shortcode execution. This is due to the software allowing users to execute an action that does not properly validate a value before running do_shortcode. This makes it possible for unauthenticated attackers to execute arbitrary shortcodes.
- CVE-2026-89274Critical
The WP Recipe Maker plugin for WordPress (versions up to and including 10.8.1) is vulnerable to arbitrary shortcode execution. The sanitize_metadata() function recursively calls do_shortcode() on recipe metadata fields, including reviewBody populated from comment content, without sanitizing shortcode tokens before execution. Unauthenticated attackers can execute arbitrary registered shortcodes on every recipe page render, disclosing data in JSON-LD metadata.
- CVE-2026-84434Critical
The Gravity Forms plugin for WordPress (versions up to and including 3.1.0.4) is vulnerable to arbitrary file upload via the upload_file function. A mismatch between validation and persistence pipelines allows hidden file upload fields to bypass extension validation. Unauthenticated attackers can upload executable files, enabling remote code execution.
- CVE-2026-93740Critical
A buffer overflow vulnerability was identified in the Totolink A3002MU (version Hh-B20211125.1046) within the formWlEncrypt function of the /boafrm/formWlEncrypt file. Manipulation of the submit-url argument allows remote exploitation. The exploit is publicly available and may be used.
- CVE-2026-93739Critical
A buffer overflow vulnerability was determined in the Totolink A3002MU (version Hh-B20211125.1046) within the formWlAc function of the /boafrm/formWlAc file. Manipulation of the submit-url argument can lead to a buffer overflow. The attack may be performed remotely, and the exploit has been publicly disclosed.
- CVE-2026-75885Critical
A flaw was found in the OpenShift console allowing unauthenticated access to the /api/devfile/ and /api/devfile/samples/ endpoints. A remote attacker can send crafted devfile payloads, leading to Server-Side Request Forgery (SSRF) and, by sending repeated large requests without a specified content length, to unbounded memory growth and Denial of Service (DoS).
- CVE-2026-93738Critical
A buffer overflow vulnerability was found in the Totolink A3002MU (version Hh-B20211125.1046) within the formSchedule function of the /boafrm/formSchedule file. Manipulation of the webpage argument results in a buffer overflow. The attack is possible remotely, and the exploit has been made public.
- CVE-2026-93839Critical
LightLLM through version 1.2.0 contains an authentication bypass vulnerability in the /pd_register WebSocket endpoint that allows unauthenticated attackers to register arbitrary nodes by supplying crafted JSON without peer address validation. Attackers can disclose full user prompts routed to their socket, trigger denial of service by replacing legitimate nodes, or make the PD Master issue requests to internal network addresses.
- CVE-2026-84082Critical
IBM Guardium Data Protection 12.2 could allow a remote attacker to execute arbitrary SQL commands due to improper neutralization of special elements used in an SQL command.
- CVE-2026-84078Critical
IBM Guardium Data Protection 12.2 is vulnerable to a missing authentication vulnerability in the LoadBalancerServlet. An unauthenticated user can access privileged load-balancer operations, potentially resulting in unauthorized actions and impact to the integrity and availability of the affected system.
Original NVD description (English source)
A vulnerability in huggingface/text-generation-inference version 3.3.6 allows unauthenticated remote attackers to exploit unbounded external image fetching during input validation in VLM mode. The issue arises when the router scans inputs for Markdown image links and performs a blocking HTTP GET request, reading the entire response body into memory and cloning it before decoding. This behavior can lead to resource exhaustion, including network bandwidth saturation, memory inflation, and CPU overutilization. The vulnerability is triggered even if the request is later rejected for exceeding token limits. The default deployment configuration, which lacks memory usage limits and authentication, exacerbates the impact, potentially crashing the host machine. The issue is resolved in version 3.3.7.

