NEWSkillSec — elevating AI Skills security from malware detection to capability auditingSkillSecLearn more →
← Back to Intel Center
SECURITY INTEL · DAILY
Link copied RSS

2026-09-17 Daily Security Intelligence

7 itemsTop severity 10.0 (Critical)CSSA 2 · CVE 2 · Poisoning 3

CSSA Exclusive Early Warning 2

CSSA Exclusive Early Warning 10.0 Critical

Corsair MCP component lacks code-execution sandbox isolation, allowing remote arbitrary code execution and host-environment disclosure

An attacker can exploit a security flaw in the run_script tool of the @corsair-dev/mcp package by constructing malicious JavaScript and passing it directly into the new Function constructor. The mechanism does not introduce a VM sandbox or membrane isolation, so the execution context shares global variables with the host process. The attacker can then access process environment variables, perform dynamic file read/write operations, and bypass read-only restrictions, ultimately causing sensitive-data disclosure, integrity damage, and loss of service availability.

Component
Corsair is an enterprise-oriented development framework for process automation and AI-agent integration. Its core architecture relies on a modular plugin system and the MCP (Model Context Protocol) protocol, aiming to connect large language models with backend business logic through standardized interfaces and to improve resource scheduling and task-orchestration efficiency.
Type
Code Injection (CWE-94)
Repo
github.com/corsairdev/corsair · Stars 11406
Remediation
  • Move code execution to the node:vm module or reuse the workflow engine's secure context; create a prototype-less isolated environment; wrap instances with a proxy membrane to prevent prototype-chain escape; and strictly forbid dynamic imports and global-object access.
CSSA Exclusive Early Warning 9.1 Critical

Floci presigned-request flaw allows authorization bypass, causing data disclosure and integrity damage

When IAM enforcement is enabled, IamEnforcementFilter only inspects the HTTP Authorization header to evaluate permission policies. Presigned URLs and presigned POST requests carry signature material in query parameters or form fields and do not set the Authorization header, so the filter skips identity-policy checks. An attacker can exploit this logic flaw by constructing a presigned request that has a valid signature but no actual write permission, bypassing explicit deny policies or permission boundaries, and then writing arbitrary objects into an S3 bucket, damaging data integrity and creating a potential information-disclosure risk.

Component
Floci is a Java-based open-source service that locally implements an AWS S3-compatible interface. Its core architecture aims to simulate Amazon S3 behavior to support local development and testing. The component uses the Quarkus framework for high-performance network interaction, provides a standardized object-storage API, and supports multi-tenant isolation plus fine-grained IAM policy evaluation. It is mainly deployed in offline or intranet environments for cloud-native application integration testing.
Type
Missing Authorization (CWE-862)
Repo
github.com/floci-io/floci · Stars 24406
Remediation
  • Extend IamEnforcementFilter at the code level so it can parse and validate authorization credentials from presigned-URL query parameters and POST form fields; ensure all authentication request types go through the unified IAM policy engine; and complete unit tests covering permission-control scenarios for each signature form.

CVE Intelligence 2

CVE-2026-73453 CVSS 10.0 Critical 2026-09-16

Unauthenticated remote code execution in Arista EOS P4Runtime leading to full administrative control

On affected platforms running Arista EOS with P4Runtime configured, an unauthenticated P4Runtime (Programming Protocol-Independent Packet Processors Runtime) client can, under certain conditions, achieve arbitrary code execution. Arista EOS disables P4Runtime by default. By constructing a malicious packet during P4Runtime session initialization, an attacker can obtain full administrative control of the compromised switch. Arista discovered this issue internally and is not aware of any malicious exploitation of this vulnerability in customer networks.

Component
Arista EOS is an enterprise-grade network operating system developed by Arista Networks. It is widely used in data-center and enterprise networks and supports advanced programmable features such as P4Runtime for flexible packet processing.
Risks
  • Full administrative control: An attacker can obtain full administrative control of a compromised switch by constructing a malicious packet and then execute arbitrary code
  • Network-infrastructure compromise: Because the vulnerability exists in a core network device, a successful exploit lets the attacker fully control network traffic, steal sensitive data, or launch further intranet attacks
  • Unauthenticated attack: The vulnerability allows an unauthenticated client to attack; the attacker needs no credentials to trigger it remotely, so the exploitation barrier is very low
Source
NVD · CVE-2026-73453
Remediation
  • Keep P4Runtime disabled unless the business explicitly requires it (the default configuration)
  • If P4Runtime must be enabled, strictly restrict source IPs and enforce network-layer access-control lists (ACLs)
  • Monitor abnormal traffic and malicious-packet characteristics during P4Runtime session initialization
CVE-2026-90038 CVSS 9.8 Critical 2026-09-16

Use-after-free in the Linux kernel NFSD module leading to remote code execution

Before Linux kernel 7.2.5 and 7.3-rc1, the NFSD (NFS server daemon) module has a use-after-free (Use-After-Free, CWE-416) security flaw caused by a concurrency-handling defect. During export-state revocation, `nfsd4_revoke_export_states()` releases `nn->client_lock` without pinning the client reference, so under a race the client may already have been freed when `revoke_one_stid()` dereferences it. An attacker can construct a specific NFS export-removal operation (such as `exportfs -u`) racing with client expiration, trigger memory corruption, and threaten system stability and security. The vulnerability affects all Linux kernel users who have not updated to a fixed version, including servers and embedded deployments. An attacker needs only network access to trigger it, with no user interaction, and can operate remotely.

Component
The Linux kernel is the core component of the Linux operating system. It manages hardware resources, process scheduling and the network protocol stack; the NFSD module provides NFS network-filesystem service.
Risks
  • From ordinary user to administrator: If an attacker can trigger a kernel-space vulnerability, they may break user-space isolation and obtain root privileges
  • Complete system control: An attacker can use this vulnerability to execute arbitrary code in kernel space, causing a crash, data disclosure, or full takeover of the server
  • No-user-interaction attack: The vulnerability is triggered directly through the network protocol stack (T1190); the attacker needs no click or interaction from the target user
Source
NVD · CVE-2026-90038
Remediation
  • Immediately upgrade the Linux kernel to 7.2.5 or 7.3-rc1 or later
  • Monitor NFS service logs for abnormal export-revocation or client-expiration race behavior
  • Restrict the network-access scope of the NFS service and allow only trusted clients

Package Poisoning 3

Package Poisoning npm 2026-09-17

kartyk-github-single-ver-pkg@>= 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
1b23219be8e89b1a4395ca0dd5e56947
Package Poisoning npm 2026-09-17

kartyk-github-token-pkg@>= 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
2674a5e3dd89e04878d924ab51fcee71
Package Poisoning npm 2026-09-17

pkg-rollback-dreed-viced-sonic-ponds@>= 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
24da0b9b6170d81f48aa593ed63e4e84
← Back to Intel CenterRSSMonitor these risks automatically with CleanSource SCA →