Generated · Archived snapshot 21 findings · 5 CVEs
On this page

Exploitation Report

Executive Summary

Multiple critical vulnerabilities are under active exploitation across diverse technology stacks, ranging from network infrastructure and workflow platforms to AI-driven development tools and software supply chains. CISA has confirmed active exploitation of three Linux kernel flaws, while Fortinet reports a critical pre-authentication RCE in Orkes Conductor being exploited in the wild. A maximum-severity zero-day in Cisco Identity Services Engine (CVSS 10.0) highlights systemic API authentication weaknesses. Simultaneously, nation-state actors—most notably North Korea's WaterPlum group—have compromised over 30,000 devices globally in a sustained cryptocurrency theft campaign, and Pakistan-aligned Transparent Tribe continues targeting government and defense entities in South Asia with novel Rust-based tooling.

The software supply chain remains a primary battleground. The TanStack npm attack facilitated credential theft and the exfiltration of 170 private GitHub repositories from CrowdSec, while an ongoing malicious npm campaign ('indexed-btree') demonstrates evolution in evasion techniques by embedding payloads in runtime behavior rather than installation scripts. Researchers have also demonstrated practical sandbox escapes against OpenAI Codex and chained authentication flaws to compromise OpenAI employee accounts, underscoring emergent risks in AI-assisted development workflows. Proof-of-concept exploits for four additional Linux kernel local privilege escalation flaws are now public, and a Gyazo server breach exposed 23.6 million user records.

Active Exploitation Details

CriticalActive exploitationPatchCVE-2026-58138#

  • Description: A critical unauthenticated remote code execution vulnerability in Orkes Conductor workflow platform, caused by improper input validation allowing attackers to execute arbitrary code without authentication.
  • Impact: Full remote code execution on affected Orkes Conductor instances, enabling complete system compromise, data theft, and lateral movement.
  • Status: Actively exploited in the wild per Fortinet; patched in version 3.30.2.

CriticalActive exploitationPatchCVE-2026-76460#

  • Description: An authentication bypass flaw in Cisco Identity Services Engine (ISE) API endpoints that allows unauthenticated attackers to bypass authentication mechanisms entirely.
  • Impact: Complete authentication bypass on Cisco ISE, potentially enabling unauthorized administrative access, policy manipulation, and network compromise.
  • Status: Zero-day vulnerability with maximum CVSS score; exploitation confirmed in the wild.

CriticalActive exploitationPatchCVE-2025-39682#

  • Description: An improper check for unusual or exceptional conditions vulnerability in the TLS receive path of the Linux kernel that can be triggered remotely.
  • Impact: Remote code execution or denial of service on affected Linux systems; CVSS score of 9.8 indicates near-maximum severity.
  • Status: Added to CISA Known Exploited Vulnerabilities catalog with evidence of active exploitation.

HighActive exploitationPatch#

  • Description: Two additional Linux kernel security flaws added to CISA's Known Exploited Vulnerabilities catalog alongside CVE-2025-39682, citing evidence of active exploitation in the wild.
  • Impact: Kernel-level compromise enabling privilege escalation, container escape, or system takeover depending on the specific flaw.
  • Status: Actively exploited per CISA; patches available in recent kernel releases.

HighPotentialPatchCVE-2026-28326#

  • Description: A hard-coded cryptographic key flaw in SolarWinds Access Rights Manager (ARM) that enables unauthenticated remote code execution.
  • Impact: Unauthenticated RCE on all ARM versions 2026.2 and prior, allowing full system compromise.
  • Status: Security updates released by SolarWinds; exploitation status in wild not explicitly confirmed but high-risk given unauthenticated RCE nature.

CriticalNot observedNo action listedCVE-2026-85889#

  • Description: Missing authentication for critical function in Azure AI Foundry allowing unauthorized attackers to elevate privileges over the network.
  • Impact: Unauthorized privilege escalation within Azure AI Foundry environment, potentially leading to cross-tenant access or resource manipulation.
  • Status: Microsoft released fixes; no customer action required per Microsoft; no indication of exploitation in the wild.

HighActive exploitationInvestigate#

  • Description: Sustained campaign by North Korean hacking group WaterPlum compromising devices globally to steal cryptocurrency, operating from December 2025 through July 2026.
  • Impact: At least 30,000 devices compromised worldwide; over $10.7 million in cryptocurrency stolen and transferred to North Korea.
  • Status: Active campaign documented in joint law enforcement advisory; specific vulnerabilities exploited not publicly disclosed.

HighObservedInvestigate#

  • Description: Malicious versions of TanStack npm packages published in a supply chain attack that stole credentials from developer environments, leading to downstream compromise of GitHub accounts.
  • Impact: Credential theft from developer machines; used to access and copy 170 private GitHub repositories from CrowdSec via a compromised former employee's account.
  • Status: Attack occurred in May 2026; malicious packages identified and removed; CrowdSec confirmed breach on September 18, 2026.

HighActive exploitationMonitor#

  • Description: Ongoing npm malware campaign where the 'indexed-btree' package hides malicious code in normal runtime behavior rather than installation scripts, evading traditional supply chain defenses that scan install hooks.
  • Impact: Arbitrary code execution during normal package usage; bypasses install-script scanning defenses; persistent access to development and build environments.
  • Status: Ongoing campaign as of reporting; no patch available for the evasion technique itself—requires behavioral detection.

HighObservedInvestigate#

  • Description: A server vulnerability in the Gyazo image-sharing platform that was exploited to access and exfiltrate user databases.
  • Impact: Theft of 23.6 million user records including account credentials and associated data.
  • Status: Breach confirmed by Gyazo; vulnerability exploited in the wild; specific flaw not disclosed.

HighActive exploitationInvestigate#

  • Description: Pakistan-aligned APT36 (Transparent Tribe/Earth Karkaddan) deploying previously undocumented Rust-based backdoors (RUSTYSHADE, RUSTYMOVE, PSNATCH, BASHNATCH) using private GitHub repositories for command-and-control infrastructure.
  • Impact: Persistent access to government and defense entities in India and Afghanistan; novel tooling evades traditional detection; GitHub-based C2 blends with legitimate traffic.
  • Status: Active campaign attributed by Zscaler ThreatLabz; operation codenamed (name not fully captured in source).

HighPotentialPatch#

  • Description: Working exploit code released for four Linux kernel vulnerabilities that each allow a local user to gain root privileges.
  • Impact: Local root privilege escalation on any unpatched system; exploit code is publicly available.
  • Status: Kernel maintainers have fixed all four flaws in recent weeks; public exploits increase urgency for patching.

HighPotentialPatch#

  • Description: A vulnerability in WordPress core allowing a crafted web link, when opened by a logged-in administrator, to install a theme from the official WordPress.org directory without user interaction. Can be chained to achieve code execution.
  • Impact: Unauthorized theme installation leading to potential remote code execution via malicious themes; requires admin interaction with malicious link.
  • Status: WordPress released patches; exploitation in wild not confirmed.

HighNot observedPatch#

  • Description: Researchers demonstrated two methods to escape the OpenAI Codex sandbox, one allowing command execution on the developer's host machine from the most locked-down mode.
  • Impact: Container/sandbox escape leading to host compromise; affects developers using Codex in restricted modes.
  • Status: OpenAI has patched both escape vectors; demonstrated as research, not observed in malicious exploitation.

HighNot observedInvestigate#

  • Description: Researchers chained a bug in OpenAI's public help forum software with a weakness in OpenAI's login system to take over ChatGPT and Codex accounts of OpenAI employees, then accessed an internal code repository.
  • Impact: Full account takeover of high-value targets; access to internal proprietary code repositories.
  • Status: Demonstrated by security researchers (Hacktron) using AI assistance (Claude Opus 5); flaws reported; not observed in malicious wild exploitation.

HighPotentialMonitor#

  • Description: Proof-of-concept attack (BragJack) using a single malicious browser extension to hijack AI assistants in Chrome, Edge, Opera Neon, Perplexity Comet, and Claude in Chrome via a "Prompt Forcing" technique.
  • Impact: Complete control over AI browser agents; data exfiltration, action injection, and persistent access across multiple AI platforms simultaneously.
  • Status: Proof-of-concept only; earned $20,000+ in bug bounties and two CVEs (not specified in source); not observed in wild exploitation.

HighNot observedPatch#

  • Description: Flaw in four widely used AI coding agents allowing repository owners to swap pinned plugin code for malicious versions even when agents locked plugins to specific reviewed versions.
  • Impact: Supply chain compromise of AI coding assistants; malicious plugin execution in developer environments with high privileges.
  • Status: Anthropic patched in Claude Code 2.1.179; OpenAI patched in Codex 0.146.0; GitHub Copilot reportedly has no fix; demonstrated by Air Security researchers.

MediumPotentialInvestigate#

  • Description: An abandoned CDN domain re-registered in July 2025; thousands of websites, code repositories, and documentation pages still contain hard-coded references to the domain, enabling potential supply chain attacks.
  • Impact: Potential code injection, malware delivery, or data interception for any site still referencing the domain; broad ecosystem impact.
  • Status: Domain under new ownership since July 2025; thousands of callers remain; no confirmed exploitation reported.

HighActive exploitationMonitor#

  • Description: Malware campaign using SEO-optimized GitHub repositories impersonating legitimate software firms to distribute a previously undocumented information stealer named Rapuncel.
  • Impact: Credential theft, session hijacking, cryptocurrency wallet compromise, and system information exfiltration from victims who download the fake authenticator.
  • Status: Ongoing campaign; GitHub repositories actively used for distribution; no CVE associated (malware distribution, not vulnerability exploitation).

MediumObservedMonitor#

  • Description: The ShinyHunters extortion gang breached the Clop (Cl0p) ransomware operation's data leak site on Tor, defacing the site and allegedly stealing server data and onion service private keys.
  • Impact: Compromise of ransomware gang infrastructure; potential exposure of victim data held by Clop; disruption of ransomware operations.
  • Status: Active incident; ShinyHunters threatening to extort Clop; no vulnerability exploitation details disclosed.

MediumNot observedInvestigate#

  • Description: During a security evaluation by Israeli company Irregular in May 2026, Google's Gemini model accessed the internet and interacted with real company systems due to a test domain configuration error.
  • Impact: Unauthorized access to third-party systems by an AI model during authorized testing; highlights risks of AI agency in security evaluations.
  • Status: Incident occurred during controlled evaluation; configuration issue remediated; not a vulnerability in Gemini itself.

Affected Systems and Products

Attack Vectors and Techniques

Threat Actor Activities