HTTPX2 is a next generation HTTP client for Python. Prior to 2.11.0, Request._prepare() in src/httpx2/httpx2/_models.py can add a body-derived Content-Length header to a request that already contains a caller-supplied Transfer-Encoding header because its setdefault() processing checks each default header independently rather than treating the two framing headers as mutually exclusive. Fixed-size byte, JSON, form, and known-length multipart bodies can therefore be serialized over HTTP/1.1 with both headers, allowing request smuggling or connection desynchronization when downstream intermediaries disagree about which framing header takes precedence. This issue is fixed in version 2.11.0.

HTTPX2 is a next generation HTTP client for Python. Prior to 2.11.0, FileField.render_headers() in src/httpx2/httpx2/_multipart.py directly interpolates attacker-controlled content_type values and custom headers from the files= three-element (filename, content, content_type) tuple and the files= four-element (filename, content, content_type, headers) tuple into multipart/form-data part headers without validating header names or values. CR or LF characters can terminate a part header, inject additional part headers, or end the part header block early, allowing a downstream multipart parser to treat attacker-supplied lines as genuine headers and potentially alter part semantics or bypass header-based checks. This issue is fixed in version 2.11.0.

HTTPX2 is a next generation HTTP client for Python. From 2.5.0 until 2.10.0, the HTTPX2 Server-Sent Events parser in src/httpx2/httpx2/_sse.py repeatedly copies and rescans buffered text in _SSELineDecoder.decode() when an attacker-controlled or compromised SSE endpoint splits one unterminated line across many response chunks. The behavior affects httpx2.Client.sse() and httpx2.AsyncClient.sse(), and the total processing work grows quadratically with the line length, allowing a crafted stream to consume excessive CPU and block a synchronous worker or asynchronous event loop. This issue is fixed in version 2.10.0.

LiteLLM is a proxy server (AI Gateway) to call LLM APIs in OpenAI (or native) format. Prior to versions 1.88.6 and 1.96.2, any authenticated LiteLLM proxy user could redirect an outbound provider call to a destination the user controls and cause the proxy to send its configured provider credentials to that destination. Request validation in litellm/proxy/auth/auth_utils.py, litellm/proxy/common_request_processing.py, litellm/proxy/health_endpoints/_health_endpoints.py, litellm/proxy/image_endpoints/endpoints.py, and litellm/proxy/litellm_pre_call_utils.py used incomplete checks that did not cover every sensitive parameter or inspect equivalent values across nested request fields, path values, and bracket-notation form data. Routing and credential parameters including api_base, base_url, model_list, fallbacks, and litellm_credential_name could therefore be applied without clearing the operator's stored key, exposing upstream provider credentials and other configured secrets and permitting server-side requests to internal services reachable by the proxy. This issue is fixed in versions 1.88.6 and 1.96.2.

libjxl before 0.12 contains an integer underflow vulnerability in the container box parser that allows remote attackers to inject arbitrary metadata by exploiting 64-bit box size truncation to size_t on 32-bit platforms. Attackers can supply a crafted JPEG XL file causing the decoder to parse attacker-controlled codestream bytes as phantom box headers, enabling injection of arbitrary metadata (Exif, XMP, IPTC, JUMBF) and potential out-of-bounds reads.

A vulnerability was identified in Sonatype Nexus Repository 3 in which two blobstore group management REST API endpoints did not correctly enforce the intended authorization check. A user granted only the nexus:blobstores:create permission could invoke these endpoints to convert an existing blobstore into a group blobstore, an action that should require the nexus:blobstores:update permission instead. This could result in unauthorized modification of blobstore configuration without administrator approval. The nexus:blobstores:create permission is a named permission that must be explicitly granted by an administrator; it is not held by default.

In affected versions of Nexus Repository 3, the script execution endpoint (POST /service/rest/v1/script/{name}/run) did not verify whether script execution had been administratively disabled. An account holding script-execution permission could continue to run previously-created scripts even after an administrator set nexus.scripts.allowCreation=false, undermining the expectation that this setting fully blocks script execution.

Nexus Repository 3 contains a sensitive information disclosure vulnerability in the capability read API. An account holding the nexus:capabilities:read privilege can retrieve the plaintext shared secret configured on a webhook capability, which is intended to be masked from all API responses. This issue affects Nexus Repository 3 versions 3.2.0 through 3.95.x, and is fixed in version 3.96.0.

An authorization flaw in the REST API repository details endpoint (GET /service/rest/v1/repositories/{repositoryName}) in Sonatype Nexus Repository 3 allowed an account holding read or browse permission on a group repository to retrieve metadata for member repositories on which it held no direct permission, by requesting the endpoint directly for the member repository name. For proxy repositories, the disclosed metadata includes the configured remote URL, which may reveal internal upstream hostnames. This includes the anonymous user if it has been granted this permission; whether the anonymous user holds this permission depends on the role and permission configuration of the specific installation.

A user account with permission to deploy artifacts to a hosted Maven repository could upload a POM file containing an oversized metadata field. This causes future attempts to list or browse that repository's components to permanently fail until an administrator repairs the underlying data. Only the targeted repository is affected; other repositories and overall server health remain unaffected.

A flaw was found in submariner. In cert-auth mode, the connection configuration is built using free-form strings from the Custom Resource Definition (CRD) without proper validation. A malicious cluster can exploit this by publishing a CableName that includes newlines and ipsec.conf directives. This allows an attacker to inject arbitrary configuration parameters or execute commands through leftupdown hooks, leading to remote code execution as root on the gateway node.

Boruta is a standalone authorization server that aims to implement OAuth 2.0 and Openid Connect up to decentralized identity specifications. Prior to version 0.10.0, Boruta logged sensitive OAuth and OpenID Connect values in business event logs. Logged values could include access tokens, refresh tokens, authorization codes, agent tokens, direct-post codes, ID tokens, VP tokens, and tokens submitted to introspection or revocation endpoints. An attacker with access to Boruta logs, log aggregation systems, or the administration log viewer could recover these credentials and use them until expiration or revocation. This issue has been patched in version 0.10.0.

PREVAIL is a Polynomial-Runtime EBPF Verifier using an Abstract Interpretation Layer. Prior to version 0.2.4, the prevail eBPF verifier accepts ALU32 ADD and SUB instructions that operate on pointer-typed registers without checking the is64 flag. Because ALU32 arithmetic zero-extends the 32-bit result, the upper half of any pointer is silently destroyed at runtime, yet prevail marks the program as verified safe. Any caller that can submit an eBPF program for verification — including unprivileged users on kernels that permit BPF program loading — can produce a program that passes verification but faults or misbehaves at runtime. This issue has been patched in version 0.2.4.

PREVAIL is a Polynomial-Runtime EBPF Verifier using an Abstract Interpretation Layer. Prior to version 0.2.4, the abstract transformer in prevail treats writes through a T_CTX-typed base register as a silent no-op: do_mem_store in src/crab/ebpf_transformer.cpp only models T_STACK stores, and the checker's T_CTX bounds arm never tests AccessType::write. An attacker can craft an eBPF program that overwrites a context field (e.g., ctx->data), reload that field typed as T_PACKET, and dereference an attacker-controlled address — and prevail will report the program as safe. This issue has been patched in version 0.2.4.

PREVAIL is a Polynomial-Runtime EBPF Verifier using an Abstract Interpretation Layer. Prior to version 0.2.4, in the Prevail eBPF verifier, EbpfTransformer::add() silently skips offset-variable updates when the destination register carries a non-singleton typeset (two or more simultaneously possible pointer types). Subsequent bounds checks use the stale offset and accept out-of-bounds memory accesses, so a crafted BPF program passes verification even though it would corrupt memory at runtime. This issue has been patched in version 0.2.4.

Joro is a web exploitation framework. Prior to version 1.1.1, Joro's default proxy mode exposes a local API on 127.0.0.1:9090 that performs no authentication and applies a wildcard CORS policy. Because plugin uploads use the CORS-safelisted multipart/form-data content type, cross-origin JavaScript on any page the operator visits can reach privileged endpoints - including uploading a native plugin and triggering a restart - directly through the operator's browser, with no preflight or credentials. Since plugins execute on load, this yields unauthenticated remote code execution as the operator's user from a single page visit. This issue has been patched in version 1.1.1.

DSpace open source software is a repository application which provides durable access to digital resources. From versions 8.0-rc1 to before 8.4, 9.0-rc1 to before 9.3, and 10-rc1 to before 10.0, a path traversal vulnerability is possible via the COAR Notify / LDN service in DSpace. The attacker MUST already have DSpace administrator credentials in order to perform the attack. When reading a file input stream of an "inbound pattern" / "template", used to generate an LDN message, the LDN class does not check for path traversal or restrict the templates to a known base path. This could allow an untrusted file from elsewhere in the file system (e.g. an export log, a bitstream path, a temporary file) to be read and interpreted as an Apache Velocity template. This issue has been patched in versions 8.4, 9.3, and 10.0.

DSpace open source software is a repository application which provides durable access to digital resources. From versions 8.0-rc1 to before 8.4, versions 9.0-rc1 to before 9.3, and version 10-rc1, Remote Code Execution (RCE) is possible via Velocity Templates used by DSpace for COAR Notify/LDN messages. This issue has been patched in versions 8.4, 9.3, and 10.0.

DSpace open source software is a repository application which provides durable access to digital resources. Prior to versions 7.6.7, 8.4, 9.3, and 10.0, the Curation Task feature allows an output path to be used by the reporter (-r parameter), typically used to stream results and status of curation task operations. It is not restricted to any particular base path, meaning that any path writable by the DSpace (often 'tomcat') user is allowed. This constitutes a Path Traversal Vulnerability in the curate script. This issue has been patched in versions 7.6.7, 8.4, 9.3, and 10.0.

DSpace open source software is a repository application which provides durable access to digital resources. Prior to versions 7.6.7, 8.4, 9.3, and 10.0, when ingesting an aggregated ORE resource by URI (using the OAI-ORE Harvester), the ORE Ingestion Crosswalk does not validate the URI scheme. This may allow for local file inclusion via malicious paths like file:///etc/passwd. The attacker MUST already have DSpace collection administrator privileges in order to perform the attack. This issue has been patched in versions 7.6.7, 8.4, 9.3, and 10.0.

Boruta is a standalone authorization server that aims to implement OAuth 2.0 and Openid Connect up to decentralized identity specifications. Prior to version 0.10.0, BorutaIdentityWeb.UserSettingsController.update/2 atomizes every key of the user-supplied request body via String.to_atom/1 before any validation. Because String.to_atom interns atoms permanently in the BEAM atom table (default cap 1,048,576 atoms; ERL_MAX_ATOMS), any authenticated end user can send PUT /users/settings with a user[<fresh-key>]=... body containing fresh keys per request and exhaust the global VM atom table. Once the table is full, the BEAM aborts with no more index entries in atom_tab and the entire OIDC server (auth, admin, gateway apps in the umbrella) crashes. The route is protected only by require_authenticated_user and a per-IP rate limit of 10 requests/second; a logged-in end user can hit it. The keys are atomized unconditionally before the downstream Accounts.update_user/6 call, so even failing updates contribute to exhaustion. This issue has been patched in version 0.10.0.

agentverus-scanner fails to analyze compiled Python bytecode files in companion code directories, allowing attackers to bypass security scanning by shipping malicious __pycache__ entries alongside benign source files. Attackers can execute arbitrary Python bytecode on import while the scanner reports a CERTIFIED verdict with high trust scores in both static and semantic analysis modes.

claude-skill-antivirus fails to analyze executable files when scanning local skill directories, reading only SKILL.md while ignoring Python source, bytecode, and other artifacts in the scripts directory. Attackers can distribute skills with malicious code in non-manifest files that receive a SAFE verdict with 100/100 trust score despite containing unanalyzed executable payloads.

Tencent AI-Infra-Guard's skill-scan component excludes compiled Python bytecode files from analysis by hardcoding __pycache__ directories and .pyc/.pyo/.pyd extensions into skip lists across multiple scanning surfaces. Attackers can distribute skills with benign Python source files alongside malicious compiled bytecode that executes on import while the scanner reports a safe verdict, enabling code execution when operators install the skill.

Astro is a web framework for content-driven websites. Prior to 7.2.4, Astro stripped a configured non-root base path from request pathnames using a string-prefix check without verifying a path-segment boundary. With base "/app", a request to "/appX/admin" resolved internally to the protected "/admin" route while middleware observed "/appX/admin" in context.url.pathname. In applications that authorize base-prefixed routes by inspecting context.url.pathname, an unauthenticated remote attacker could bypass pathname-based middleware authorization and reach protected routes. This issue is fixed in version 7.2.4.