CSSA Exclusive Early Warning10.0 Critical
lmdeploy P2P connector remote code execution via deserialization of untrusted data
The vulnerability stems from engine_conn.py applying no authentication on ZMQ TCP sockets and directly calling recv_pyobj on raw network bytes. That method effectively runs pickle.loads, so an attacker can send a malicious Pickle-serialized payload to the exposed port. Because type checks occur after deserialization, the defense fails, enabling arbitrary code execution on the engine node and leading to full system compromise and lateral movement risk.
Component
lmdeploy is a distributed deployment framework focused on accelerating large language model inference. Its core architecture uses a disaggregated serv…
Type
Deserialization of Untrusted Data (CWE-502)
Repo
Remediation- It is recommended to replace unsafe serialization with secure protocols such as JSON or MsgPack combined with strict schema validation, and to enable ZMQ CURVE encryption and authentication so that only authorized clients can connect and exchange data.
CSSA Exclusive Early Warning9.6 Critical
Megabasterd remote code execution and data disclosure risk from deserialization flaws and unsafe input handling
An attacker can exploit Java deserialization during Settings import or database session restore, chaining the commons-collections gadget on the classpath for remote code execution. Unvalidated file-path decryption allows arbitrary file writes; an unauthenticated local streaming server enables SSRF; weak key derivation and plaintext storage increase credential disclosure risk; maliciously crafted PBKDF2 iteration parameters can cause denial of service; and a misconfigured global authenticator lets third-party servers capture proxy credentials.
Component
Megabasterd is a Java-based MEGA cloud storage desktop client for file sync, encrypted storage and offline access. Its architecture integrates local d…
Type
Deserialization of Untrusted Data (CWE-502)
Repo
Remediation- Enforce strict object deserialization filtering and migrate sensitive storage to JSON; canonicalize decrypted file paths to prevent traversal; add random token validation and Host header checks for local service interfaces; use a strong KDF such as Argon2id with AES-GCM for data at rest; and bound external numeric parameters to prevent resource exhaustion.
CSSA Exclusive Early Warning10.0 Critical
cc-connect unauthorized Webhook endpoint remote code execution via command injection leading to full system compromise
The vulnerability lies in authentication and command execution on the /hook endpoint. When webhook.token is empty, authenticate returns true and bypasses access control. An attacker can send an unauthenticated POST with a malicious exec field that is passed unsanitized into exec.CommandContext running /bin/sh -c, executing arbitrary OS commands in the service process context and fully compromising the server with sensitive data disclosure risk.
Component
cc-connect is an integrated DevOps automation connector whose core architecture receives external CI/CD pipeline events and triggers an internal workf…
Type
OS Command Injection (CWE-78)
Repo
Remediation- Fixes should require a non-empty high-entropy token for authentication, remove or strictly limit direct OS command execution, apply strict allowlist filtering and character escaping on user input, and run the service with least privilege to limit blast radius.
CSSA Exclusive Early Warning10.0 Critical
cc-connect management interface full server compromise via missing authentication combined with command injection
An attacker can abuse authentication bypass caused by a default empty token, sending malicious requests to the unprotected cron creation endpoint so that a user-controlled exec parameter is passed directly into shellExecCommand. The defect stems from missing input filtering and authentication, allowing a remote attacker without credentials to run arbitrary OS commands on the host, then steal sensitive data, plant backdoors or move laterally to other internal assets.
Component
cc-connect is middleware for automated process management. Its core architecture includes a management service, cron scheduling engine and execution e…
Type
OS Command Injection (CWE-78)
Repo
Remediation- Fix authentication so an empty token denies access, enforce a strict command-parameter allowlist, avoid concatenating shell commands directly, and use a firewall to limit network access to management interfaces, eliminating injection risk at the root.
CVE-2026-77956CVSS 10.0 Critical2026-08-31
ash_ai code injection leading to remote code execution
In ash_ai prior to version 1.0.0, the AshAi.Actions.Prompt module improperly handles prompt content, creating a code injection (CWE-94) security defect. The component evaluates prompt content with EEx.eval_string/2; when prompt action text includes request data, attacker-controlled text is compiled and run as an EEx template (Elixir source). Content such as <%= System.cmd(...) %> executes on the server before any model request is issued. The vulnerability affects all ash_ai users not updated to 1.0.0, including early versions from 0.1.0. An attacker only needs to send a crafted request to an exposed prompt action, with no authentication required, and can control the system remotely.
Component
ash_ai is an open-source library under ash-project for integrating AI capabilities into Elixir applications, providing handling of AI prompts and actions.
Risks
- Complete system control: An attacker can execute arbitrary Elixir code on the server and fully control the affected application and underlying system
- Unauthenticated attack: Exploitation requires no authentication; access to an exposed prompt action interface is enough
- Remote code execution: An attacker can run system commands before model requests are issued, causing data disclosure, service disruption or lateral movement
Source
Remediation- Immediately upgrade ash_ai to 1.0.0 or later
- Review whether function-supplied prompt content is evaluated as EEx and ensure only statically configured templates are evaluated
- Enforce strict access control and input validation on exposed AI interfaces
CVE-2026-82542CVSS 10.0 Critical2026-08-30
Boa Web Server remote buffer overflow leading to remote code execution
In Tenda HG10 firmware 300001138, the formIPv6Routing function in /boaform/admin/formIPv6Routing of the Boa Web Server component has a security defect. It does not perform strict bounds checking on the destNet parameter, resulting in improper restriction of operations within the bounds of a memory buffer (CWE-119) and a buffer overflow. An attacker can trigger it remotely over the network with a crafted request; exploit code has been publicly disclosed and may be used in attacks. The vulnerability affects all Tenda HG10 devices running the affected firmware. No authentication is required, attack complexity is low, and the system may be fully compromised.
Component
Boa Web Server is a lightweight single-task web server commonly used in embedded network devices such as routers to handle HTTP requests.
Risks
- Complete system control: An attacker can execute arbitrary code on the victim device and obtain full control
- Remote unauthenticated attack: The attack vector is network-based with no user interaction or privileges required
- High-impact consequences: CVSS 3.1 scoring shows high impact on confidentiality, integrity and availability, with changed scope
Source
Remediation- Deploy a firewall at the network boundary to restrict external access to device management interfaces such as /boaform/admin/
- Monitor network traffic for anomalous requests targeting the formIPv6Routing interface
CVE-2026-81096CVSS 10.0 Critical2026-08-28
ToolUniverse sandbox escape leading to remote code execution
In ToolUniverse prior to version 1.3.0, the executor behind the python_code_executor tool (in python_executor_tool.py) has a sandbox escape flaw. Although it checks forbidden attribute names and calls, it retains the attribute-lookup builtin and does not block dunder attributes reached via string lookup or allowed modules. An attacker can walk from a literal's class to its base and enumerate subclasses to obtain references to the process and subprocess modules. Per-call arguments also let the caller widen the import allowlist before checks run. Meanwhile, the HTTP and MCP servers in http_api_server.py and smcp_server.py bind to all interfaces, enable debug mode and require no authentication, so any caller who can reach the port can execute code as the server process. The vulnerability affects all ToolUniverse users not updated to 1.3.0. An unauthenticated attacker can trigger it remotely and fully control the server.
Component
ToolUniverse is an open-source project that runs caller-supplied Python code in a sandbox, commonly used for automation tasks or code-execution services.
Risks
- Complete system control: An attacker can execute arbitrary code on the server as the server process and obtain full control
- Unauthenticated remote attack: Because the server binds to all interfaces without authentication, any attacker who can reach the port may attack
- Privilege escalation: Sandbox escape can grant access to process and subprocess modules and thus system command execution
Source
Remediation- Immediately upgrade ToolUniverse to 1.3.0 or later, which adds Bearer Token authentication, defaults the bind address to loopback, and strengthens attribute checks
- If an immediate upgrade is not possible, bind the server to 127.0.0.1 only and disable debug mode
- Apply network ACLs so only trusted IPs can reach the relevant ports
- Monitor the server process for anomalous behavior such as unexpected child processes or outbound connections
CVE-2026-82222CVSS 10.0 Critical2026-08-28
GiveWP deserialization of untrusted data leading to object injection
In GiveWP prior to version 4.16.7.1, the data processing module lacks strict validation of input data, creating a deserialization of untrusted data (CWE-502) security defect. An attacker can craft malicious serialized data to trigger object injection, which may lead to remote code execution or sensitive information disclosure. The GiveWP plugin has a deserialization vulnerability when handling certain requests; an attacker can launch it remotely over the network with no user interaction and low complexity, seriously threatening confidentiality, integrity and availability. It affects all GiveWP users not updated to 4.16.7.1, including all WordPress sites deploying the plugin. An attacker only needs to send a crafted request to exploit it and remotely control the affected system.
Component
GiveWP is an open-source WordPress donation plugin by Liquid Web / StellarWP, widely used by nonprofits and charities for online fundraising.
Risks
- From ordinary user to administrator: If the victim site is misconfigured, an attacker may escalate privileges and obtain administrator control
- Complete system control: An attacker can execute arbitrary code on the victim server and, depending on privileges, install malware, view/modify/delete data or create fully privileged accounts
- No user interaction required: Via remote network requests (T1190), an attacker can trigger the vulnerability without luring users to click, enabling automated attacks
Source
Remediation- Immediately upgrade GiveWP to 4.16.7.1 or later
- Deploy a web application firewall (WAF) to block anomalous deserialization requests
- Monitor server logs to identify and block suspicious object injection attempts
CVE-2026-82456CVSS 10.0 Critical2026-08-29
ArgoCD MCP unauthorized access leading to remote code execution
In argocd-mcp version 0.8.0, when ARGOCD_API_TOKEN is configured, the HTTP transport module binds incorrectly, creating a binding to an unrestricted IP address (CWE-1327) security defect. The component binds to all network interfaces and accepts MCP sessions without caller credentials. An attacker who can reach the listener can use the operator-stored token to invoke the full tool set, creating applications, requesting syncs and modifying Argo CD resources. The vulnerability affects all unpatched argocd-mcp 0.8.0 deployments. Network reachability alone is enough; no extra authentication is required for remote control.
Component
ArgoCD MCP is a component in the Argo CD ecosystem that provides application sync and resource management interfaces via the Model Context Protocol (MCP).
Risks
- From ordinary user to administrator: If an attacker can reach the exposed service port, they can use the stored administrator token to fully control the Argo CD cluster
- Complete system control: An attacker can create malicious apps in the Kubernetes cluster, sync malicious configs or modify existing resources, potentially leading to arbitrary code execution or data disclosure inside the cluster
- No user interaction required: An attacker needs no victim action and can trigger the vulnerability by directly accessing the exposed MCP interface over the network
Source
Remediation- Restrict the argocd-mcp listen address to localhost or trusted internal interfaces only
- Enforce strict Network Policies to block unauthorized traffic to the MCP service port
- Regularly rotate ARGOCD_API_TOKEN and monitor anomalous tool-call logs
Package Poisoningnpm2026-08-31
@fidzzhost/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
ec831478fb3f57ae29784a8088383cb7
Package Poisoningnpm2026-08-31
brat-codex@>= 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
766972ec1cd9ed5211cc4e5bb74186d0
Package Poisoningnpm2026-08-31
cloufcrxz@>= 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
d2e8343812e2b352ae31c1e078c040bb
Package Poisoningnpm2026-08-31
h2-codex@>= 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
bc0a5bb001e1627c8e0bf1add533b501
Package Poisoningnpm2026-08-31
h3-codex@>= 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
dd8f9d0ddbed98af8622a04ac6d8fb45
Package Poisoningnpm2026-08-31
originaldevelopmentstelemetry@>= 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
f47cb15b2b9ed3841dcb57ff80037e65
Package Poisoningnpm2026-08-31
real-browser-plus@>= 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
aa13ddce28c13c4bd8bc794834b1ee38
Package Poisoningnpm2026-08-31
@lucideproxy/svg@>= 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
41aa7b3c270b0418aaf0e1b3b00b8242
Package Poisoningnpm2026-08-31
eslint-prettier-js@>= 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
e6cc8be8bdfc81bc1176b94db9440679
Package Poisoningnpm2026-08-31
redis-cookie-server@>= 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
e0c31a232420f2ee401fa2129fede823
Package Poisoningnpm2026-08-28
tailwindcss-3d-animate@>= 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
e0622ab006233acc967ddfc961ec4b41
Package Poisoningnpm2026-08-28
@hd-team/app-dnpkg-eight@>= 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
a87fae3ce665cfecea23d03ec5f8b38c
Package Poisoningnpm2026-08-28
@hd-team/app-dnpkg-prod@>= 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
d27b64311ae8ee507d0e1c91de990bf0
Package Poisoningnpm2026-08-28
@hd-team/app-dnpkg-ten@>= 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
42c93882e00123b1861f35ac15517d84
Package Poisoningnpm2026-08-28
@hd-team/app-dnpkg-test@>= 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
9411ad3eee7bcbb49ab811eed3c54f8a
Package Poisoningnpm2026-08-28
@hd-team/app-dnpkg-three@>= 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
605991ca3d159de945b470cd95718900
Package Poisoningnpm2026-08-28
@hd-team/app-impkg-prod@>= 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
4ecb6a827300f071df4a70a3d0a34d24
Package Poisoningnpm2026-08-28
@hd-team/app-impkg-test@>= 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
600f45b886a555369709037085092b9d
Package Poisoningnpm2026-08-28
hydration-ui-dlx@>= 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
6724cd52359fd8c3ff9b5268d82179e8
Package Poisoningnpm2026-08-28
svelte-ui-dlx@>= 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
bb91cac05ddbbba23d8e8a36b3244109
Package Poisoningnpm2026-08-29
amplitude-session-replay@45.0.0 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
152136aee53e4a8bf7da2ff2ef5195af
Package Poisoningnpm2026-08-29
aws-blog@45.0.2 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
b7f7c4d2cc21fcb2bd49b7c5b962c029
Package Poisoningnpm2026-08-29
intuit-authz@55.0.0 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
5f9191dd90dc9a78c6bed542ca88e8fc
Package Poisoningnpm2026-08-29
analytics-web-client@30.0.0 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
703876921fc588c5a5c33c7137c05a19
Package Poisoningnpm2026-08-29
aura-instrumentation@45.0.0 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
0df4275a872ed6831db1f24ce93c28b2
Package Poisoningnpm2026-08-29
confluence-create-content@30.0.0 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
c773aeb2f61a5a5f9ca8fc01fc66973c
Package Poisoningnpm2026-08-29
jira-projects-backbone@45.0.0 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
dfcb32ac019aa665381710c2ba5d1eac
Package Poisoningnpm2026-08-29
mkt-ui-library@45.0.0 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
0ad694a0cf5df4be1d03f8da5ba816b2
Package Poisoningnpm2026-08-29
oit-lib-oracle-util@45.0.0 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
7d8c9ae6c13bb75d35d7b1250ceb10cf
Package Poisoningnpm2026-08-29
prm-bundles@45.0.0 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
d520961855be6c4b71060ee11be1152e
Package Poisoningnpm2026-08-30
htps-provider@>= 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
d7deae3c4aebe6dc846d230928019aef
Package Poisoningnpm2026-08-30
manager-thedate@>= 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
5c6369b790f20c3cca3a488c0ad426eb
Package Poisoningnpm2026-08-30
nuvyra-marketplace-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
efdc4d5219756a4725646966e3bdc085
Package Poisoningnpm2026-08-30
node-net-pool@>= 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
69d5a1f717f9a562c8ebccfcbb4164af
Package Poisoningnpm2026-08-30
gclassroom@>= 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
e443ecf915581c5181d768b945e16c8f
Package Poisoningnpm2026-08-30
opiumbest@>= 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
dd9bb27499b5ee1d7208362acd789ae1
Package Poisoningnpm2026-08-30
quesoeducation@>= 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
779c913f559afc556b72ce75f1a95fc3
