LiteLLM version 1.107.0 was compromised in a supply chain attack affecting PyPI packages and Docker images. The malicious release includes Python startup files that execute automatically, stealing cloud credentials, private keys, browser data, and AI development tool authentication tokens. The attack used a lookalike repository in the release workflow and targets credentials for AWS, Azure, Google Cloud, GitHub CLI, SSH keys, npm/PyPI, and AI tools such as Claude, Gemini, and Cursor. Stolen data is encrypted and sent to a command-and-control server. A loader enables remote Python code execution. The package has approximately 23 million weekly downloads, indicating a significant potential impact.
IBM ContextForge MCP Gateway 0.1.0 through 1.0.5 is vulnerable to a server-side request Forgery (SSRF) that allows authenticated users with gateway creation privileges to perform arbitrary HTTP POST requests to internal network resources and cloud metadata endpoints. The vulnerability occurs during OAuth-enabled gateway registration when the system fails to validate the oauth_config.token_url parameter. When registering a gateway with OAuth client credentials authentication, an attacker can supply a malicious token_url pointing to internal services, cloud metadata endpoints (AWS IMDSv1, GCP, Azure), or localhost-bound services. The application immediately performs an HTTP POST request to the attacker-controlled URL with form-encoded credentials in the body. If the response cannot be parsed as a valid OAuth token response, up to 256 characters of the response body are reflected back to the attacker in the HTTP 502 error message. This vulnerability bypasses existing SSRF protections that are applied to the gateway.url field but not to OAuth configuration parameters.
Bulletin ID: 2026-054-AWS Scope: AWS Content Type: Important (requires attention) Publication Date: 07/14/2026 13:00 PM PDT Description: AWS HealthLake MCP Server (awslabs.healthlake-mcp-server) is a Model Context Protocol server that enables AI assistants to interact with AWS HealthLake FHIR datastores. We identified CVE-2026-15643, a server-side request forgery in the pagination handling component in AWS awslabs.healthlake-mcp-server before 0.0.14 on all platforms might allow a remote authenticated user to exfiltrate AWS temporary security credentials to an arbitrary endpoint via a crafted next_token parameter. The server does not validate that pagination URLs point back to the expected HealthLake endpoint, allowing an actor to redirect subsequent requests to an actor-controlled server. Impacted versions: < 0.0.14 Please refer to the article below for the most up-to-date and complete information related to this AWS Security Bulletin.
Bulletin ID: 2026-052-AWS Scope: AWS Content Type: Important (requires attention) Publication Date: 07/06/2026 13:45 PM PDT Description: Amazon mcp-gateway-registry is an open-source gateway and registry for Model Context Protocol (MCP) servers, providing centralized discovery, authentication/authorization, and proxying of MCP tools for AI agents. We identified CVE-2026-14471, an issue in the metrics-service retention policy management component where a caller-supplied table_name value is interpolated into SQL statements in identifier position without proper neutralization. An authenticated remote user is able to supply a crafted table_name value to execute arbitrary SQL queries against the metrics database. This allows the user to read stored data (including API key material) and to delete or alter stored data. Impacted versions: >=1.0.3 AND <=1.0.12 Please refer to the article below for the most up-to-date and complete information related to this AWS Security Bulletin.
Bulletin ID: 2026-051-AWS Scope: AWS Content Type: Important (requires attention) Publication Date: 07/01/2026 12:45 PM PDT Description: The AWS Advanced JDBC Wrapper is an open-source JDBC driver wrapper that extends a JDBC driver to enable Amazon Aurora and AWS Cloud features such as failover handling and caching. We identified CVE-2026-14265, an issue in the RemoteQueryCachePlugin of the AWS Advanced JDBC Wrapper. When this plugin is enabled, query results read from the shared Redis/Valkey cache are deserialized without class filtering. An actor with write access to the shared cache infrastructure could insert a crafted serialized Java object that, when read by an application, results in execution of arbitrary code on the application server. Impacted versions: >=3.3.0 AND <=4.0.0 Please refer to the article below for the most up-to-date and complete information related to this AWS Security Bulletin.
CVE-2026-110812 is a vulnerability in Amazon Neptune's implementation of Apache TinkerPop's Gremlin graph-import functionality. It allows an authenticated user with query permissions to submit crafted input that leads to unsafe processing of untrusted data within the query engine. This could result in unintended activity within the query engine environment. AWS recommends upgrading to fixed engine versions to mitigate this issue.
The FBI has arrested a suspected member of the ShinyHunters extortion group linked to a recent breach of FBI systems via a third-party vendor platform. The group exploited an alleged Oracle PeopleSoft zero-day vulnerability to access FBI systems and later moved into FBI-managed AWS GovCloud infrastructure, stealing 2-3TB of sensitive data including employee personal and medical information. Multiple arrests and law enforcement actions have targeted ShinyHunters members internationally. The FBI attributes the breach to a failure by a third-party contractor to install a security update. ShinyHunters is known for data theft and extortion campaigns against various organizations worldwide, often leveraging stolen credentials and phishing attacks to access cloud SaaS platforms.
CVE-2026-107322 is an OS command injection vulnerability in the Amazon Agent Plugins for AWS, specifically in the databases-on-aws plugin versions 1.0.0 through 1.7.0. This plugin extends AI coding agents with AWS workflows and includes database helper scripts. The vulnerability arises from an incomplete input validation that allows a remote unauthenticated actor to supply crafted content. This crafted input can lead to operating system command execution on the host if the agent invokes the local helper with the malicious database command. The commands execute with the permissions of the helper process. This issue does not affect the Aurora DSQL service or bypass database privileges. A fix is available in version 1.7.1.
A path traversal vulnerability exists in the Backstage plugin catalog backend modules for Azure Blob Storage and AWS S3. Insufficient validation of storage object paths allows catalog descriptors to be read from outside the intended storage boundaries, limited to locations accessible by the backend's configured credentials. This issue affects versions prior to 0.3.21 for the Azure module, 0.4.27 for the AWS module, and 0.7.18 for the backend defaults. The vulnerability has a low severity with a CVSS score of 3.0. Patches are available, and recommended mitigations include restricting blob/object creation or renaming to trusted principals, scoping credentials and network access, or disabling affected catalog providers if restrictions cannot be enforced.
The dedh-devops-automation package on PyPI contains malicious code that executes during installation. It deploys a rogue AWS Lambda function designed to collect credentials from AWS Secret Manager and exfiltrate sensitive data. The package overrides the install command in setup.py to perform these malicious actions. Multiple versions of the package are affected, including 0.0.1, 5.0.0, 5.6.0, 5.6.1, 5.6.999, 5.8.0, 5.8.1, and 5.9.0. This campaign is identified as having clear malicious intent, targeting cloud tokens and credentials.
CVE-2026-104002 is a vulnerability in the data masking utility of Powertools for AWS Lambda (Python) that causes fail-open error handling. This flaw can allow unauthorized actors to access sensitive field values that should have been masked. The issue affects versions from 3.6.0 through 3.34.0 inclusive. AWS has issued a security bulletin to raise awareness of this vulnerability.
Improper neutralization of argument delimiters in the volume handling component in AWS EFS CSI Driver (aws-efs-csi-driver) v3.1.0 through v3.4.2 might allow remote authenticated users with PersistentVolume creation permissions to inject arbitrary mount options via comma-separated values in the mounttargetipmap volumeAttribute.
To remediate this issue, users should upgrade to version v3.5.0 or later.
Unverified ownership in Barman snapshot backup deletion allows a principal who can write the backup catalog to cause Barman to delete unrelated cloud snapshots. When a snapshot backup is deleted, either explicitly or by retention policy enforcement, Barman reads the snapshot identifiers from the backup.info file and passes them to the cloud provider's delete API using Barman's own credentials, without verifying that the snapshots belong to that backup. An attacker who can overwrite backup.info but lacks snapshot delete permissions can substitute the identifiers of other snapshots, causing Barman to delete any snapshot its cloud identity can reach on AWS, Microsoft Azure, or Google Cloud. Exploitation requires a deployment where the principal that writes the backup catalog is separate from the identity Barman uses to delete snapshots. Barman versions from 3.4.0 (Google Cloud), 3.6.0 (Azure), and 3.7.0 (AWS) up to and including 3.20.0 are affected. The issue is fixed in Barman 3.20.1.
---
_-= Per source details. Do not edit below this line.=-_
## Source: amazon-inspector (6457aa73ec15f68e06005977033aaba349467d40e3d406f0810684a43b74fcb5)
The package's preinstall script and its main entry (index.js) both perform identical credential-harvesting: they query the EC2 Instance Metadata Service (IMDSv2) at 169.254.169.254 for the identity document and IAM role security credentials, enumerate the entire process.env, and recursively read /var/run/secrets/ (Kubernetes service-account tokens and mounted secrets). The collected data is POSTed to a hardcoded attacker-controlled host at hrnmn.dd.h4x.tv (paths /save_instance_info_hook from preinstall and /save_instance_info_index from index.js). The destination is unrelated to any legitimate package purpose, there is no user configuration, and the behavior fires automatically on `npm install` via the preinstall lifecycle hook as well as on any `require()`/`import` of the package. The harvested material — temporary AWS IAM credentials, cloud instance identity, arbitrary environment variables (which routinely contain API keys, tokens, and database credentials in CI/build environments), and Kubernetes service-account tokens — grants immediate cross-service access to the installer's cloud and cluster resources.
## Source: ghsa-malware (07386b2d1e2efa71fd38c591953684c11302ad77f1c7943bd542b71fd83184e3)
Any computer that has this package installed or running should be considered fully compromised. All secrets and keys stored on that computer should be rotated immediately from a different computer. The package should be removed, but as full control of the computer may have been given to an outside entity, there is no guarantee that removing the package will remove all malicious software resulting from installing it.
The npm package @hrmony/benefit-essenszuschuss (version 3.99.0) contains malicious code that, upon import, harvests cloud and container secrets from the host environment. It retrieves AWS IAM role credentials via the Instance Metadata Service (IMDSv2), collects environment variables, and reads Kubernetes secrets from the filesystem. This sensitive data is then exfiltrated to a hardcoded attacker-controlled endpoint. The package mimics an internal HR utility and includes an unpinned dependency, facilitating a dependency confusion attack. Systems with this package installed should be considered fully compromised.
The npm package '@hrmony/pladddform-cli' version 40.14.3 contains malicious code that exfiltrates sensitive cloud and Kubernetes credentials to an attacker-controlled endpoint. Upon import, it sends a beacon, retrieves AWS IAM role credentials and instance metadata, and collects Kubernetes secrets and environment variables, posting all data to a suspicious external domain. This enables immediate cloud account and cluster takeover. The package name is a typosquatting variant of a legitimate package, increasing the risk of accidental installation. Systems with this package installed should be considered fully compromised.
The npm package @hrmony/pladddform-partner-registry is a malicious typosquatting package that exfiltrates cloud and container secrets upon import. It harvests AWS EC2 metadata, IAM role credentials, environment variables, Kubernetes secrets, and other sensitive data, sending them to an attacker-controlled endpoint. The package provides no legitimate functionality and is part of a coordinated typosquatting and dependency confusion campaign. Systems with this package installed should be considered fully compromised.
The npm package @hrmony/api-gateway-service-config (version 0.99.0) contains malicious code that, upon import, steals cloud IAM credentials and Kubernetes secrets from the host. It fetches AWS IMDSv2 tokens, instance metadata, environment variables, and secrets, then exfiltrates this data to a hardcoded attacker endpoint. The package also depends on an unpublished scoped package, indicating a dependency confusion attack. Installing or running this package compromises the host, requiring immediate secret rotation and package removal.
CVE-2026-55160 is a Server-Side Request Forgery (SSRF) vulnerability in the stringer-rss Stringer application, a self-hosted RSS reader. Before a specific patch commit, authenticated users could make the server send arbitrary HTTP/HTTPS requests to internal networks, localhost, and cloud metadata endpoints such as AWS IMDS. This vulnerability allows low-privileged users, especially when self-service signup is enabled, to scan internal services or potentially steal cloud IAM credentials. The issue has been patched in commit 75cb095.