The falcon-rust crate's discrete-Gaussian sampler uses a 32-bit fractional precision approximation for its exponential function, significantly less precise than the Falcon reference implementation. This reduced precision causes the sampler to deviate from the ideal discrete Gaussian distribution required for Falcon's security proof, potentially leaking secret information through signature outputs. The flaw was introduced in a regression before version 0.1.3 and results in failure to reproduce Falcon's known-answer test vectors. The maintainers acknowledged the precision loss but loosened test tolerances instead of restoring precision. A fix is available in version 0.2.0.
An SQL Injection vulnerability exists in Bolt CMS's ListFormatHelper component, allowing authenticated users with upload permissions to execute arbitrary SQL queries via the /async/select-options endpoint when the general/caching/list_format setting is enabled. This vulnerability is fixed in Bolt version 6.1.4.
Bolt versions prior to 6.1.5 contain a server-side request forgery (SSRF) vulnerability in the /bolt/async/upload-url and /bolt/async/embed endpoints. These endpoints validate only the initially submitted URL but do not re-validate URLs after following redirects, allowing authenticated users to bypass SSRF protections and make requests to internal or cloud metadata addresses. The vulnerability has been patched in Bolt 6.1.5.
CVE-2026-133406 is a medium severity vulnerability in aoiflux mutant versions 2.3.0, 2.4.0, and 2.5.0. It involves uncontrolled memory allocation caused by manipulation of the Count argument in the Forensic Parser component, specifically in the builtin/hive_builtins.go file. The vulnerability can be exploited locally. Upgrading to version 2.6.0 resolves the issue.
CVE-2026-77379 is an information disclosure vulnerability in Bolt CMS prior to version 6.1.4. It arises from a missing access-control filter in the public Relation API, allowing unauthenticated users to access metadata of unpublished, draft, and viewless content via REST and GraphQL endpoints. This could expose sensitive content metadata without authentication. The issue has been fixed in Bolt version 6.1.4.
CVE-2026-77377 is a Server-Side Request Forgery (SSRF) vulnerability in Bolt CMS affecting authenticated backend endpoints that fetch user-supplied URLs without protection against internal network targets. This flaw allows an attacker with authenticated access to cause the server to make unauthorized requests, potentially exposing internal resources. The vulnerability has been assigned a CVSS 3.1 score of 5.2 (medium severity). It is fixed in Bolt version 6.1.4.
An open redirect vulnerability in Bolt's LoginFormAuthenticator allows attackers to redirect users to arbitrary attacker-controlled domains after login success or failure. This can facilitate phishing attacks and potential token theft. The issue affects Bolt versions prior to 6.1.4 and has been patched in version 6.1.4.
Daptin contains a broken access control vulnerability (CVE-2026-77369) in its generic YJS collaboration WebSocket endpoint. This endpoint allows any authenticated user to read and write content of any file-typed column in any table without proper authorization checks. The flaw arises because the generic endpoint only verifies the presence of a user session but does not enforce per-resource permission checks, unlike the secured sibling endpoint. This enables unauthorized access and modification of admin-only or otherwise restricted data via the collaboration sync mechanism.
Daptin versions prior to 0.12.30 contain an unauthenticated SQL injection vulnerability via the `sort` query parameter on all entity list endpoints. The vulnerability arises because the `sort` parameter is concatenated directly into SQL identifiers without validation, allowing attackers to inject arbitrary SQL. This can lead to denial of service or unauthorized data extraction without authentication.
Jupyter MCP Server versions prior to 1.0.3 contain a vulnerability where custom management routes do not enforce the same browser-origin protections as the main /mcp endpoint. This allows a malicious website to perform a DNS rebinding attack, sending crafted requests to internal MCP server routes such as /api/connect and /api/stop. These routes accept forged Host and Origin headers, enabling an attacker to replace the active Jupyter backend configuration and control kernel execution. The vulnerability is fixed in version 1.0.3.
The Jupyter MCP Server in standalone streamable-http mode enforces authentication via MCP_TOKEN on the /mcp endpoint but fails to enforce this token on custom management routes such as /api/connect and /api/stop. This allows unauthenticated attackers to reset the server context, redirect the Jupyter backend to attacker-controlled infrastructure, start kernel clients, and stop notebook kernels without authentication. The vulnerability is fixed in version 1.0.3.
flyto-core versions prior to 2.26.8 contain a Server-Side Request Forgery (SSRF) guard bypass vulnerability due to DNS rebinding. The SSRF guard resolves the hostname once and validates the IPs, but the actual connection performs a separate DNS resolution without pinning the validated IP. This allows an attacker controlling DNS to flip the resolved IP from a public to a private/internal address between the guard's check and the connection, bypassing protections and potentially accessing internal services such as cloud metadata endpoints.
Flyto2 Core's http.paginate function validates only the initial base URL before following server-supplied pagination URLs from Link headers or cursor/offset URLs without revalidating them. This allows an attacker to craft a first-page response that points the 'next' URL to an internal address, causing a second-order SSRF. The internal response body is aggregated and returned to the caller, exposing sensitive internal information. Versions up to 2.26.6 are affected, with a fix available in version 2.26.8.
Flyto2 Core versions up to 2.26.6 contain a Server-Side Request Forgery (SSRF) vulnerability due to a time-of-check to time-of-use (TOCTOU) race condition in the SSRF guard. The function validate_url_ssrf resolves a hostname to an IP address and checks if it is private, but then discards the IP and returns the hostname. At connection time, aiohttp performs a fresh DNS resolution, allowing an attacker controlling DNS to alternate the resolved IP between a public and an internal IP, bypassing the SSRF guard. This enables unauthorized internal network fetches. The issue is fixed in version 2.26.8 by pinning the validated IP address for connection.
Flyto-core versions prior to 2.26.8 contain a critical arbitrary file write vulnerability in several data.* and file.* modules due to incomplete path validation. These modules fail to enforce sandbox confinement, allowing an attacker with module execution privileges to write files to arbitrary absolute paths on the host. This can lead to overwriting application code, user shell configurations, cron jobs, or SSH authorized keys, resulting in remote code execution and full host compromise. The issue is an incomplete fix of a prior critical vulnerability (GHSA-2956), where path validation was applied only to a subset of modules. The vulnerability is rated critical with a CVSS score of 10. A fix is available in version 2.26.8.
django-docutils versions prior to 0.30.0 render reStructuredText (RST) using Docutils without secure defaults, allowing attacker-controlled RST to exploit unsafe Docutils directives. This can lead to local file disclosure, server-side request forgery (SSRF), and stored cross-site scripting (XSS) via raw HTML injection. The vulnerability affects applications that render untrusted or attacker-influenced RST content. The issue is fixed in version 0.30.0 by disabling unsafe directives and sanitizing output by default.
jane-php versions prior to 7.12.0 have a vulnerability in their $ref resolution during code generation. The software resolves external $ref targets using file_get_contents without restricting the URI scheme or path, allowing attacker-controlled OpenAPI documents to trigger server-side request forgery (SSRF) and local file inclusion (LFI) via path traversal. This can lead to arbitrary HTTP requests and local file reads at build or CI time, with the fetched content inlined into generated code.
Refitter versions prior to 2.1.2 contain a critical vulnerability (CVE-2026-63344) where unescaped OpenAPI content-type keys are directly embedded into Refit attribute strings. This allows an attacker to inject arbitrary code that executes at assembly load time via a crafted content-type key in the OpenAPI specification. The vulnerability has a CVSS 4.0 score of 9.3 and affects Refitter 2.0.0 and earlier. A fix is available in version 2.1.2.
Refitter versions prior to 2.1.2 contain a vulnerability where unescaped OpenAPI header parameter names are emitted directly into Refit attribute strings. This allows an attacker to inject code that executes at assembly load time via a crafted header parameter name in the OpenAPI specification. The flaw was verified on Refitter 2.0.0 with .NET 8 and can lead to code execution at application startup.
Refitter versions prior to 2.1.2 allow arbitrary code execution via unescaped OpenAPI path strings embedded directly into Refit attribute strings. An attacker can craft a single-line path that closes the attribute and interface context, injects a file-scoped class with a .NET ModuleInitializer method, and reopens namespaces/interfaces to maintain valid code. This injected code executes at assembly load, enabling code execution at application startup. The vulnerability affects Refitter 2.0.0 and earlier and is fixed in version 2.1.2.