Unauthenticated broadcast in Dynamo's KVBM merger and dropped tenant-isolation identifiers leak sensitive tokens and mix multi-tenant data
The flaw comes from the KV event merger binding itself to the 0.0.0.0 interface and dropping the cache_namespace field during data processing. An attacker needs no authentication to sniff cleartext token_ids from the network. Because the tenant salt is also missing, data from different tenants is incorrectly indexed into a global unsalted key space, breaking multi-tenant isolation and leaking sensitive context.
Component
Dynamo is a distributed KV-cache management component optimized for large-language-model inference. Its core architecture depends on a vLLM event-stream subscription mechanism, and it is designed to improve memory utilization through KV-block merging and deduplication, support cache acceleration and resource scheduling in multi-tenant environments, and provide high-performance key-value storage routing and management.
Restrict the network listen address to the local loopback interface, ensure the event-transformation logic fully preserves tenant-isolation identifiers, and enforce a strict access-control policy to prevent unauthorized remote connections.
CSSA Exclusive Early Warning
10.0 Critical
CR/LF injection in YDB HTTP/2 pseudo-headers causes request splitting and damages data integrity
An attacker can exploit a flaw in the HTTP/2 CompleteHeaderBlock function, which does not strictly filter carriage-return and line-feed characters in pseudo-header values, and craft a request packet that contains illegal control characters. When StreamToIncomingRequest processes the request, it inserts the unsanitized method, path, and authority headers directly into a synthesized HTTP request. Ordinary headers are checked, but that check cannot block injection through pseudo-headers. The parser then sees a tampered request line or extra injected headers, causing HTTP request/response splitting. The affected system faces a high-severity risk of sensitive-data leakage and tampered request logic.
Component
YDB is a distributed cloud-native database platform. Its architecture is designed for high availability and horizontal scaling. A core component is the HTTP/2-based network communication module, which handles efficient data transfer and command exchange between clients and servers, and uses stream processing to optimize concurrent connection management so that large-scale reads and writes stay low-latency and the system remains stable.
Apply strict input validation and sanitization to pseudo-header values before building the HTTP representation, and reject any data that contains carriage-return or line-feed characters. Also harden the HTTP/2 frame parser so that every header field goes through a uniform safety check before it is merged, preventing a malicious payload from reaching the lower application layer.
CVE Intelligence
2
CVE-2026-53998
CVSS
10.0 Critical
2026-11-17
Phoenix GraphQL F-string template injection leads to remote code execution
In Phoenix versions 13.0.0 through 15.7.0, the `applyChatTemplate` GraphQL query is flawed: it passes an attacker-controlled F-string template to `FStringTemplateFormatter._format`, and `DotDict.__getattr__` does not block traversal through Python private attributes. A carefully crafted field can walk internal object attributes (such as `__class__`, `__init__`, and `__globals__`) and read process environment variables and internal module state. In a default install, the endpoint requires no authentication; even when authentication is enabled, any account can reach it. Disclosure of PHOENIX_SECRET may allow forged administrative session tokens and decryption of credentials protected by keys derived from that secret. This issue was fixed in version 15.7.0.
Component
Phoenix is an AI observability and evaluation platform for monitoring and assessing the performance and behavior of AI applications.
Risks
Sensitive information disclosure: An attacker can read process environment variables and internal module state, and obtain critical configuration data
Privilege escalation and session hijacking: Disclosure of PHOENIX_SECRET can allow forged administrative session tokens, so the attacker can impersonate an administrator
Credential decryption: An attacker can decrypt credentials protected by keys derived from PHOENIX_SECRET and move further into the system
Unauthenticated remote exploitation: In the default configuration, an attacker can trigger the flaw remotely with no authentication, so the barrier to exploitation is very low
Upgrade Phoenix to version 15.7.0 or later immediately to fix this flaw
If an immediate upgrade is not possible, restrict network access to the GraphQL endpoint and allow only trusted IP addresses
Enable and enforce a strong authentication mechanism, and rotate the PHOENIX_SECRET key on a regular schedule
Monitor GraphQL query logs for unusual F-string template-injection attempts
CVE-2026-111559
CVSS
10.0 Critical
2026-11-16
Remote code execution in the Batflat CMS /contact endpoint allows arbitrary PHP code execution
In Batflat CMS, the /contact endpoint does not filter incoming data, an improper neutralization of special elements used in a template engine (CWE-1336). A remote, unauthenticated attacker can send a POST request containing a crafted message and execute arbitrary PHP code as the www-data user. Every system that uses Batflat CMS is affected. Because the project is no longer maintained and the flaw has not been fixed, the risk remains. An attacker can trigger it remotely with no authentication and take control of the server directly.
Component
Batflat CMS is a content management system that provides website content management and publishing, and supports handling user-interaction data through endpoints such as /contact.
Risks
Complete system control: An attacker can use this flaw to execute arbitrary PHP code on the server, and read, modify, or delete server files as the www-data user
Unauthenticated remote attack: The attacker needs no credentials or user interaction, only a crafted POST request, to trigger the flaw
Persistent backdoor: An attacker can upload a webshell or malicious script, establish a long-lived access channel, and move further into the internal network
Stop using Batflat CMS immediately and migrate to an actively maintained, trustworthy alternative CMS
If an immediate migration is not possible, block malicious POST requests to the /contact endpoint at the web application firewall (WAF)
Limit the server's ability to parse untrusted external input, and monitor abnormal process behavior under the www-data user
Package Poisoning
10
Package Poisoning
npm2026-11-17
spencer-baileys@1.0.4
flagged as malicious
This version was found communicating with a malicious domain and executing malicious commands. Audit your dependencies and pin safe versions immediately.
MD5
735478f8bebf5bcf590754e91e9b0bef
Package Poisoning
npm2026-11-17
sauruspro-bail@5.0.0
flagged as malicious
This version was found communicating with a malicious domain and executing malicious commands. Audit your dependencies and pin safe versions immediately.
MD5
be87f2281e031391aa4f292047d1564a
Package Poisoning
npm2026-11-17
sauruskyah-baileys@>= 0
flagged as malicious
This version was found communicating with a malicious domain and executing malicious commands. Audit your dependencies and pin safe versions immediately.
MD5
83bfc0587e8eef26ef48dc35239809f6
Package Poisoning
npm2026-11-17
wanzy@>= 0
flagged as malicious
This version was found communicating with a malicious domain and executing malicious commands. Audit your dependencies and pin safe versions immediately.
MD5
c5071801064dd7d9f492170b17c5dbfa
Package Poisoning
npm2026-11-17
@cygnus-finance/sdk@>= 0
flagged as malicious
This version was found communicating with a malicious domain and executing malicious commands. Audit your dependencies and pin safe versions immediately.
MD5
03e521bd259342963d6440af01d491da
Package Poisoning
npm2026-11-17
analytics-skd@>= 0
flagged as malicious
This version was found communicating with a malicious domain and executing malicious commands. Audit your dependencies and pin safe versions immediately.
MD5
5235e6ea9ab2206b50dbd68767a432c4
Package Poisoning
npm2026-11-17
@finzzcloud/baileys@>= 0
flagged as malicious
This version was found communicating with a malicious domain and executing malicious commands. Audit your dependencies and pin safe versions immediately.
MD5
b37b7d70e2b0a348677aaae32cb60726
Package Poisoning
npm2026-11-17
@devforgespr/postcss-animation-toolkit@>= 0
flagged as malicious
This version was found communicating with a malicious domain and executing malicious commands. Audit your dependencies and pin safe versions immediately.
MD5
fde69d923d55ad4315cf9bf169e7b909
Package Poisoning
npm2026-11-17
gugoddiscordapi@>= 0
flagged as malicious
This version was found communicating with a malicious domain and executing malicious commands. Audit your dependencies and pin safe versions immediately.
MD5
69c2edab377fd24223a95ca7bd3f31c4
Package Poisoning
npm2026-11-17
jito-bundle-service@>= 0
flagged as malicious
This version was found communicating with a malicious domain and executing malicious commands. Audit your dependencies and pin safe versions immediately.