Vulnerabilities

SGLang contains an unauthenticated RCE in /load_lora_adapter_from_tensors via bypass of SafeUnpickler’s incomplete denylist, allowing arbitrary command execution through crafted base64-encoded pickle payloads.

A stored cross-site scripting (XSS) vulnerability exists in the `POST /api/prompts/share` endpoint of parisneo/lollms (latest version). The endpoint stores attacker-controlled `prompt_content` into `DBDirectMessage.content` without server-side sanitization. When a victim opens the direct message (DM) thread, the message is rendered by the DM UI through `MessageContentRenderer`, which uses `v-html` to insert rendered HTML into the DOM. The frontend sanitizer, which is regex-based, fails to comprehensively sanitize attacker-controlled HTML, allowing malicious payloads to execute in the victim's browser context. This vulnerability enables any authenticated user to send a malicious prompt-share message to another user's inbox, leading to arbitrary JavaScript execution, authenticated actions as the victim, exposure of same-origin application data, and potential account takeover.

SGLang uses an expert-parallel backup subsystem that exposes a ZeroMQ PULL socket on a routable network interface that does not contain authentication or deserialization safeguards, allowing an attacker to provide a malicious pickle file that results in unauthenticated remote code execution when the feature is enabled and the service is reachable over the network.

A vulnerability was determined in sgl-project SGLang up to 0.5.11. Affected by this vulnerability is the function data_hash of the component Cache Handler. This manipulation causes denial of service. The attack is restricted to local execution. A high degree of complexity is needed for the attack. The exploitation appears to be difficult. The exploit has been publicly disclosed and may be utilized. The pull request to fix this issue awaits acceptance.

SGLangs multimodal generation runtime is vulnerable to unauthenticated remote code execution when the --enable-custom-logit-processor option is enabled, as Python objects loaded via dill.loads() will be deserialized without validation.

SGLang's reranking endpoint (/v1/rerank) achieves Remote Code Execution (RCE) when a model file containing a malcious tokenizer.chat_template is loaded, as the Jinja2 chat templates are rendered using an unsandboxed jinja2.Environment().

A Cross-site Scripting (XSS) vulnerability was identified in the `from_dict` method of the `AppLollmsMessage` class in parisneo/lollms prior to version 2.2.0. The vulnerability arises from the lack of sanitization or HTML encoding of the `content` field when deserializing user-provided data. This allows an attacker to inject malicious HTML or JavaScript payloads, which can be executed in the context of another user's browser. Exploitation of this vulnerability can lead to account takeover, session hijacking, or wormable attacks.

A Stored Cross-Site Scripting (XSS) vulnerability was identified in the social feature of parisneo/lollms, affecting the latest version prior to 2.2.0. The vulnerability exists in the `create_post` function within `backend/routers/social/__init__.py`, where user-provided content is directly assigned to the `DBPost` model without sanitization. This allows attackers to inject and store malicious JavaScript, which is executed in the browsers of users viewing the Home Feed, including administrators. This can lead to account takeover, session hijacking, and wormable attacks. The issue is resolved in version 2.2.0.

In parisneo/lollms version 2.1.0, the application's session management is vulnerable to improper access control due to the use of a weak secret key for signing JSON Web Tokens (JWT). This vulnerability allows an attacker to perform an offline brute-force attack to recover the secret key. Once the secret key is obtained, the attacker can forge administrative tokens by modifying the JWT payload and resigning it with the cracked secret. This enables unauthorized users to escalate privileges, impersonate the administrator, and gain access to restricted endpoints. The issue is resolved in version 2.2.0.

A vulnerability in parisneo/lollms, up to and including version 2.2.0, allows unauthenticated users to upload and process files through the `/api/files/extract-text` endpoint. This endpoint does not enforce authentication, unlike other file-related endpoints, and lacks the `Depends(get_current_active_user)` dependency. This issue can lead to denial of service (DoS) through resource exhaustion, information disclosure, and violation of the application's documented security policies.

Sakai is a Collaboration and Learning Environment (CLE). In versions 23.0 through 23.4 and 25.0 through 25.1, group titles and description can contain cross-site scripting scripts. The patch is included in releases 25.2 and 23.5. As a workaround, one can check the SAKAI_SITE_GROUP table for titles and descriptions that contain this info.

LoLLMs Web UI (lollms-webui), the web interface for Lord of Large Language and Multimodal Systems, contains a critical unauthenticated server-side request forgery (SSRF) flaw in its POST /api/proxy endpoint. An attacker with network access to the interface can send a crafted request that forces the server to issue arbitrary GET requests to attacker-chosen destinations. This enables access to internal-only services, scanning of the host's local network, and exfiltration of sensitive cloud instance metadata such as AWS or GCP IAM credentials. All known existing versions of lollms-webui are affected, and as of publication no patched release is available, so any deployed instance—particularly internet-exposed or cloud-hosted ones—is at risk. No confirmed in-the-wild exploitation is known, but a public security advisory exists and EPSS assigns an elevated 21.6% probability of exploitation within 30 days (97th percentile).

Do: No patched version is available yet, so mitigate by restricting access to the /api/proxy endpoint (e.g., via reverse-proxy authentication or ACLs) and limiting which instances are internet-exposed; on cloud-hosted deployments, block or restrict access to the instance metadata service (e.g., 169.254.169.254) to prevent IAM credential theft. Review server and cloud logs for unexpected outbound GET requests or metadata-service hits indicating exploitation, and upgrade as soon as a fixed lollms-webui release is published.

SGLang's multimodal generation module is vulnerable to unauthenticated remote code execution through the ZMQ broker, which deserializes untrusted data using pickle.loads() without authentication.

A user enumeration vulnerability exists in FormaLMS 4.1.18 and below in the password recovery functionality accessible via the /lostpwd endpoint. The application returns different error messages for valid and invalid usernames allowing an unauthenticated attacker to determine which usernames are registered in the system through observable response discrepancy.

Sakai is a Collaboration and Learning Environment. Prior to versions 23.5 and 25.0, EncryptionUtilityServiceImpl initialized an AES256TextEncryptor password (serverSecretKey) using RandomStringUtils with the default java.util.Random. java.util.Random is a non‑cryptographic PRNG and can be predicted from limited state/seed information (e.g., start time window), substantially reducing the effective search space of the generated key. An attacker who can obtain ciphertexts (e.g., exported or at‑rest strings protected by this service) and approximate the PRNG seed can feasibly reconstruct the serverSecretKey and decrypt affected data. SAK-49866 is patched in Sakai 23.5, 25.0, and trunk.

Improper Neutralization of Special Elements used in an SQL Command ('SQL Injection') vulnerability in chrisbadgett LifterLMS lifterlms allows SQL Injection.This issue affects LifterLMS: from n/a through <= 8.0.6.

The LifterLMS WordPress plugin before 8.0.1 does not sanitise and escape a parameter before outputting it back in the page, leading to a Reflected Cross-Site Scripting which could be used against high privilege users such as admin.

A path traversal vulnerability exists in the `install` and `uninstall` API endpoints of parisneo/lollms-webui version V12 (Strawberry). This vulnerability allows attackers to create or delete directories with arbitrary paths on the system. The issue arises due to insufficient sanitization of user-supplied input, which can be exploited to traverse directories outside the intended path.

The LifterLMS – WP LMS for eLearning, Online Courses, & Quizzes plugin for WordPress is vulnerable to Unauthenticated Post Trashing due to a missing capability check on the delete_access_plan function and the related AJAX calls in all versions up to, and including, 8.0.1. This makes it possible for unauthenticated attackers to change status to "Trash" for every published post, therefore limiting the availability of the website's content.

The LifterLMS – WP LMS for eLearning, Online Courses, & Quizzes plugin for WordPress is vulnerable to arbitrary post deletion due to a missing capability check on the 'llms_delete_cert' action in all versions up to, and including, 7.8.5. This makes it possible for authenticated attackers, with Subscriber-level access and above, to delete arbitrary posts.

parisneo/lollms-webui version 9.6 is vulnerable to Cross-Site Scripting (XSS) and Open Redirect due to inadequate input validation and processing of SVG files during the upload process. The XSS vulnerability allows attackers to embed malicious JavaScript code within SVG files, which is executed upon rendering, leading to potential credential theft and unauthorized data access. The Open Redirect vulnerability arises from insufficient URL validation within SVG files, enabling attackers to redirect users to malicious websites, thereby exposing them to phishing attacks, malware distribution, and reputation damage. These vulnerabilities are present in the application's functionality to send files to the AI module.

A CORS misconfiguration in parisneo/lollms-webui prior to version 10 allows attackers to steal sensitive information such as logs, browser sessions, and settings containing private API keys from other services. This vulnerability can also enable attackers to perform actions on behalf of a user, such as deleting a project or sending a message. The issue impacts the confidentiality and integrity of the information.

A vulnerability in the discussion image upload function of the Lollms application, version v9.9, allows for the uploading of SVG files. Due to incomplete filtering in the sanitize_svg function, this can lead to cross-site scripting (XSS) vulnerabilities, which in turn pose a risk of remote code execution. The sanitize_svg function only removes script elements and 'on*' event attributes, but does not account for other potential vectors for XSS within SVG files. This vulnerability can be exploited when authorized users access a malicious URL containing the crafted SVG file.

Sakai is a Collaboration and Learning Environment. Starting in version 23.0 and prior to version 23.2, kernel users created with type roleview can log in as a normal user. This can result in illegal access being granted to the system. Version 23.3 fixes this vulnerability.

A vulnerability in parisneo/lollms-webui version 9.8 allows for a Denial of Service (DOS) attack when uploading an audio file. If an attacker appends a large number of characters to the end of a multipart boundary, the system will continuously process each character, rendering lollms-webui inaccessible. This issue is exacerbated by the lack of Cross-Site Request Forgery (CSRF) protection, enabling remote exploitation. The vulnerability leads to service disruption, resource exhaustion, and extended downtime.

A path traversal vulnerability exists in the api open_personality_folder endpoint of parisneo/lollms-webui. This vulnerability allows an attacker to read any folder in the personality_folder on the victim's computer, even though sanitize_path is set. The issue arises due to improper sanitization of the personality_folder parameter, which can be exploited to traverse directories and access arbitrary files.

A path traversal vulnerability exists in the parisneo/lollms-webui repository, specifically in the `lollms_file_system.py` file. The functions `add_rag_database`, `toggle_mount_rag_database`, and `vectorize_folder` do not implement security measures such as `sanitize_path_from_endpoint` or `sanitize_path`. This allows an attacker to perform vectorize operations on `.sqlite` files in any directory on the victim's computer, potentially installing multiple packages and causing a crash.

A Local File Inclusion vulnerability exists in parisneo/lollms-webui versions below v9.8. The vulnerability is due to unverified path concatenation in the `serve_js` function in `app.py`, which allows attackers to perform path traversal attacks. This can lead to unauthorized access to arbitrary files on the server, potentially exposing sensitive information such as private SSH keys, configuration files, and source code.