
Cloud Threat Highlights: H1 2026
Wiz's H1 2026 cloud threat report documents a 60% jump in significant incidents driven by a doubling of supply-chain attacks (TeamPCP's npm/PyPI/GitHub Actions credential-harvesting spree, its open-sourced Shai-Hulud worm spawning Miasma/Hades/Megalodon/IronWorm follow-ons, North Korea's axios and @mastra trojanizations) and by growing attacks on AI infrastructure, including exposed unauthenticated MCP servers, LiteLLM/Dify/Langflow/n8n/Ollama vulnerabilities, and cryptominers deployed via a Langflow RCE. It also profiles JINX-0163, a cloud-native extortion group targeting non-human identities (service accounts/IAM roles) across AWS/Azure/GCP/Okta/Snowflake linked to the "FulcrumSec" extortion brand, and closes with concrete detection techniques for stolen access keys, Kubernetes VM Run Command abuse, GCP org-policy tampering, and CloudTrail user-agent/boto3 credential-provenance mismatches.
What it is
Four detection techniques and one supply-chain mitigation from Wiz's H1 2026 cloud threat report. The existing summary names these signals ("stolen access keys, Kubernetes VM Run Command abuse, GCP org-policy tampering, CloudTrail user-agent/boto3 mismatches") but does not carry the actual detection logic, which is the reason to open the page.
What to do
Hidden persistence via CreateAccessKey. After planting a backdoor admin, attackers profile IAM users, find an active administrator, and add a second access key on that legitimate account — persistence that hides behind a real employee rather than a new user. Detect: monitor CreateAccessKey events where the calling principal is not the account owner.
Kubernetes node escape via VM Run Command. Attackers with access to managed Kubernetes call the VM Run Command API against the underlying nodes for root code execution outside the Kubernetes layer, bypassing container controls. Detect: monitor VM Run Command calls in Azure Activity Logs.
GCP org-policy tampering. With compromised service-account keys, an attacker adds a foreign domain to the org's allowedPolicyMemberDomains, grants an external account editor on a project, then removes both the grant and the policy change within thirty minutes — transient and rarely audited. Detect: SetOrgPolicy calls on allowedPolicyMemberDomains from identities that are not known IAM admins, especially when followed by SetIamPolicy in the same session.
Stolen instance-role credentials (near-zero false positive). Boto3 reads the local OS at runtime, so credentials stolen from a Linux EC2 instance and run from an attacker's Windows machine surface in CloudTrail with os/windows in the user-agent — an OS mismatch against the instance. Additionally, the Boto3 feature-flag combination m/e (CREDENTIALS_CODE) means credentials were loaded explicitly into a script rather than sourced from the metadata service, which legitimate EC2 workloads almost never do. The combination of an OS mismatch on an instance role plus that flag is described as a near-zero-false-positive signal.
Supply-chain cooldown. Because community focus has shortened how long malicious packages stay downloadable, Wiz recommends a "cooldown" policy: do not install packages published less than 24 hours ago.
Caveats
Wiz assesses only with moderate confidence that FulcrumSec is JINX-0163's public identity, and cannot rule out an Extortion-as-a-Service model with separate actors per role.
Organisations named here
- Midnight Neptune (UNC1069)
- Shai-Hulud
Technologies named here
- npm
- Wiz
- Boto3
- PyPI
- CloudTrail
- Composer
- VS Code extensions
- Azure Activity Logs
- Jenkins plugins
- AUR
- GitHub Actions
- axios
- @mastra packages
- AntV
- @redhat-cloud-services npm namespace
- LiteLLM
- Dify
- Langflow
- n8n
- Ollama
- GCP
- Azure
- AWS
- Okta
- Snowflake
- Managed Kubernetes services
- Amazon Linux 2023
- Windows Server 2025
- Model Context Protocol
- mastra
- Red Hat
Techniques named here
- Software supply-chain compromise via poisoned packages
- OIDC token extraction from CI runner memory
- Maintainer account takeover and malicious package publishing
- Wiper deployment triggered by token revocation
- Branch protection stripping and forced malicious commits
- Credential harvesting and cross-group credential reuse
- Worm-based supply chain propagation
- eBPF rootkit process/network concealment
- Tor-routed C2 communications
- Non-human identity targeting for lateral movement
- Cryptomining deployment in model-serving environments
- Prompt-injection-driven coding agent abuse
- Unauthenticated MCP endpoint exploitation
- IAM persistence via CreateAccessKey on existing admin
- Container escape via VM Run Command API
- Org policy manipulation for transient external access
- Instance role credential theft with OS/user-agent mismatch
Threat actors named here
- TeamPCP
- Lapsus$
- North Korea's Midnight Neptune
- JINX-0163
- FulcrumSec
- Shai-Hulud
- Megalodon
- Miasma
- Hades
- IronWorm
- DPRK
Related in Supply Chain


