A security vulnerability in iperf3 allows the server to accept unbounded peer-controlled JSON parameters, which can lead to remote denial of service via resource exhaustion. This issue is tracked as CVE-2026-71217 and affects Red Hat Enterprise Linux 9 and its variants. Red Hat has issued a security advisory and provided updated packages to address this vulnerability.
Server-side request forgery (SSRF) in the /har/test endpoint in QD 20220208 through 20250803. Fetcher.build_request() in libs/fetcher.py constructs an httpclient.HTTPRequest from user-supplied JSON without validating URL scheme, host, or IP range. The /har/test handler does not require authentication, enabling unauthenticated remote attackers to force the QD server to send arbitrary HTTP requests to internal network resources and cloud metadata endpoints. validate_cert is set to False, disabling TLS verification.
A security update for Red Hat Hardened Images RPMs addresses multiple vulnerabilities in the grafana13.2 package, including CVE-2026-75899, which involves a flaw in the fast-uri Node.js library. This flaw allows double decoding of percent escapes in hostnames, enabling Server-Side Request Forgery (SSRF) and bypass of host-based security policies. The update includes fixed RPMs for aarch64 and x86_64 architectures. No explicit patch details are provided, and mitigation options are limited or unavailable according to Red Hat.
The npm package 'search-reservation' version 55.0.0 contains malicious code delivered via a self-referential dependency pointing to an external URL. Installing this package causes npm to fetch and execute code from a remote server without integrity checks, potentially compromising the host system. The shipped index.js is inert, making the malicious payload delivered entirely through the dependency URL. Any system with this package installed should be considered fully compromised, and secrets should be rotated immediately.
MariaDB Connector/Node.js versions prior to 3.2.4, 3.3.3, 3.4.6, and 3.5.3 can expose account passwords in cleartext during PAM dialog authentication over insecure transport. This occurs because the connector does not enforce secure transport for the PAM password authentication dialog, allowing a hostile or on-path server to trigger password disclosure over plain TCP. The issue is mitigated by using properly verified TLS or local Unix sockets. The vulnerability has a CVSS score of 5.9 (medium severity) and is fixed in the specified versions.
CVE-2026-77184 is a medium severity SQL injection vulnerability in MongoDB Connector for BI version 2.1.0. It arises because the description text of a collection's JSON schema validator is included in the comment text of DDL statements generated by SHOW CREATE without properly escaping backslash characters. A user with permission to modify the schema validator can embed additional SQL code in this output. If this generated SQL statement is later executed against a SQL server, the injected SQL runs with the privileges of that session.
MariaDB Connector/Node.js versions prior to 3.3.3, 3.4.6, and 3.5.3 have an improper certificate validation vulnerability. When SSL is enabled without a pinned CA or server certificate, the connector may send credentials before completing certificate fingerprint validation. This allows a man-in-the-middle attacker to present a self-signed certificate, capture the database password, and authenticate directly even though the connection is later closed. The issue is fixed in versions 3.3.3, 3.4.6, and 3.5.3.
SvelteKit versions 2.49.0 through 2.52.1 with experimental remote functions and form enabled have a CPU exhaustion vulnerability in form deserialization. An attacker can send malformed form data to cause the server to become unresponsive, resulting in denial of service. This issue is fixed in version 2.52.2.
SvelteKit versions 2.49.0 through 2.52.1 with experimental remote functions and form enabled have a memory exhaustion vulnerability in remote form deserialization. Malformed form data can cause excessive memory allocation, crashing the server and causing denial of service. This issue is fixed in version 2.52.2.
This security update for vim addresses multiple vulnerabilities including buffer overflows, use-after-free, out-of-bounds access, and arbitrary code execution issues. The update upgrades vim to version 9.2.0957 and fixes several security flaws that could lead to denial of service, process crashes, or potential arbitrary code execution. These vulnerabilities affect various components such as the socket server, JSON decoding, spell file loading, text property handling, popup opacity, vimball record files, shell command handling, and netrw menu entries.
Trilium is an open-source hierarchical note-taking application. In versions prior to 0.104.0, the default-on "Safe import" filter fails to neutralize the shareTemplate relation because that relation is not marked as dangerous, allowing an attacker-supplied import archive to plant a server-side template that leads to remote code execution. The relation is omitted from the built-in list of dangerous attributes, so unlike other code-loading relations it is not disabled on import, and when the victim later publishes the imported note the public share renderer feeds the linked EJS code note's raw bytes into ejs.render, which compiles them in the server's Node process. An unauthenticated request to the shared note then executes the attacker's JavaScript with full access to require, process, the filesystem, and the network. This issue is fixed in version 0.104.0.
CVE-2026-59278 is a Server-Side Request Forgery (SSRF) vulnerability in Spring for Apache Kafka. The issue arises because JsonKafkaHeaderMapper and DefaultKafkaHeaderMapper include java.net in their default trusted packages list. This default configuration allows an external Kafka producer to inject a java.net.InetAddress type via the spring_json_header_types message header. The vulnerability affects multiple versions of Spring for Apache Kafka, including 4.1.0, 4.0.0 through 4.0.6, 3.0.0 through 3.3.16, 2.9.0 through 2.9.14, and 2.8.12 and earlier. The CVSS score is 6.5, indicating a medium severity level.
CVE-2026-47879 is a high-severity Server-Side Request Forgery (SSRF) vulnerability in Spring Cloud Gateway's JsonToGrpcGatewayFilterFactory. It allows attackers with high privileges to specify arbitrary Spring Resource locations for proto descriptor definitions, potentially leading to complete confidentiality and integrity compromise. The vulnerability affects multiple versions of Spring Cloud Gateway up to 5.0.2 and earlier. No official patch or remediation guidance is currently available.
The pybitjs package on PyPI contains malicious code that executes during installation and every Python interpreter startup. It uses obfuscated JavaScript to fetch attacker-controlled payloads via a command-and-control mechanism leveraging Ethereum blockchain transactions to dynamically determine the C2 server IP. The payload downloads and executes further malicious code, consistent with campaigns attributed to the Lazarus APT group. This allows arbitrary code execution on the host system of anyone installing or running Python with this package present.
CVE-2026-80348 is a high-severity vulnerability in TarsCloud TarsWeb versions up to and including 3.0.16. It involves missing authorization checks in four methods of the PatchController.js, allowing authenticated users with limited roles to perform unauthorized actions across applications. These actions include uploading and deploying packages to any server, downloading or deleting packages from any application, and changing the default deployed package. The vulnerability arises because some methods do not call the AuthService to enforce per-application role restrictions, enabling privilege escalation within the TarsWeb console.
CVE-2026-77310 is a Server-Side Request Forgery (SSRF) vulnerability in the jackson-databind library's deserialization of java.net.InetAddress. Unlike the previously fixed InetSocketAddress deserialization which avoids DNS resolution, the InetAddress deserialization still performs eager DNS lookups on attacker-controlled input. This can lead to DNS-based SSRF allowing out-of-band interaction or internal host enumeration via DNS queries during JSON deserialization.
The openai-pr-reviewer npm package version 1.0.0 contains malicious code that executes automatically during installation. Its preinstall script collects sensitive system information including hostname, username, home directory, DNS servers, current working directory, package.json contents, and contents of /etc/passwd and /etc/hosts. This data is exfiltrated via HTTPS to a hardcoded external server. Installation requires no user interaction, making compromise silent and immediate. Systems with this package installed should be considered fully compromised, and all secrets and keys on those systems should be rotated from a different, trusted machine.
Xinference loads models with Hugging Face remote code execution unconditionally enabled, and before version 2.12.0 exposes no setting to disable it. Six loader call sites pass trust_remote_code=True as a literal or as an unconditional default: RerankModel._get_tokenizer in xinference/model/rerank/core.py, SentenceTransformerRerankModel.load in xinference/model/rerank/sentence_transformers/core.py, SentenceTransformerEmbeddingModel.load in xinference/model/embedding/sentence_transformers/core.py, FlagEmbeddingModel.load in xinference/model/embedding/flag/core.py, and two sites in xinference/model/llm/transformers/core.py where PytorchModel._sanitize_model_config and PytorchModel._get_components default the value to True. Because a caller with model launch access can register a model whose type is unknown and supply an arbitrary model path, the server reaches _auto_detect_type and then AutoTokenizer.from_pretrained, which imports and executes Python declared by the model directory's own tokenizer_config.json auto_map, running attacker-supplied code with the privileges of the worker process. Version 2.12.0 gates every site behind allow_trust_remote_code and the XINFERENCE_TRUST_REMOTE_CODE setting, permitting remote code only for bundled built-in models.