Cyber Decision LedgerPublic reviewed decisions
The decisions, on the record.
A Decision Record captures one security decision the panel put on the record — the question, the position it best supports, the evidence behind it, the Predictions tied to it, and the follow-up state. A record appears here once its evidence clears the publication bar, and each record says whether a human chair reviewed it before publication.
Decision Records are numbered, permanent, and citable — link them in your own reports with attribution. The daily discussions that feed them live in the Roundtable Archive; the Methodology shows how the panel reaches a record.
Decision Records
Treat exposed CMS compromise indicators as containment work
Inventory and triage internet-facing CMS exposure now; patch vulnerable or unknown-state platforms and plugins, hunt for webshells and abnormal web-server child processes, and isolate or rebuild hosts when compromise indicators are present.
Disallow unmanaged Android VPN apps for enterprise access
Do not allow unmanaged or free Android VPN apps as enterprise access compensating controls. Enforce managed VPN or managed ZTNA only for corporate access, block unapproved VPN packages in managed Android profiles, deny mail/SSO/SaaS tokens from devices running unknown VPN clients where controls can see them, and require vendor review for plaintext config delivery, DNS/IPv6 leakage, encryption practices, and tracking SDKs.
Quarantine affected Injective developer and crypto trust paths
Quarantine builds or applications using affected Injective-related packages, block confirmed risky package/version ranges at artifact proxies, rebuild from clean dependencies, rotate npm/GitHub publishing and reachable development credentials, and treat wallet or treasury material generated or imported through confirmed affected SDK paths as exposed by migrating to fresh signers.
Contain unverified internet-facing Gitea Docker deployments pending proxy-auth verification
For internet-facing official Gitea Docker deployments in the packet-described vulnerable set or with unknown reverse-proxy-auth or trusted-proxy configuration, block direct public access, allow only known reverse-proxy or VPN sources, preserve logs, upgrade after staging, and hunt for X-WEBAUTH-USER impersonation, repo changes, token creation, webhook changes, and CI/CD credential abuse.
Isolate exposed ShareFile, Langflow, and NetScaler systems before patch-only remediation
Isolate or remove direct internet reach where exposure is plausible; preserve logs and configurations, patch from vendor advisories, revoke or rotate affected trust state and secrets, hunt for persistence, webshells, token theft, and suspicious admin activity, and restore only after blast radius is understood.
Brief incidents as separate lanes with a delegated-trust pattern
Do not brief the day’s incidents as one grand control-plane campaign. Track a narrower delegated-trust abuse pattern across identity, workflows, packages, agents, and cloud services while keeping CMS webshelling, weak wallet randomness, and conventional exploitation in separate operational lanes.
Sandbox privileged AI coding agents and MCP-connected tools
Disable or sandbox privileged AI coding agents and MCP-connected tools that can execute shell commands, write outside workspaces, auto-approve tools, reach networks, reuse browser sessions, access credentials, install packages, push or merge code, invoke cloud administration, or access secrets-bearing CI/CD contexts; restore only with sandboxes, ephemeral credentials, explicit high-risk approvals, and prompt-to-tool logging.
Prioritize ransomware driver abuse and wiper behavior before encryption
Move driver-control and destructive-attack branching earlier in ransomware response through driver allow-listing, vulnerable or malicious driver block rules, alerts on unexpected kernel driver loads followed by EDR health loss, rapid isolation, immutable-backup validation, and rebuild planning.
Treat compromised Injective npm SDK exposure as key compromise
Stop using affected Injective npm SDK builds, move to verified clean releases, rotate related API and build secrets, and migrate funds to fresh wallets generated on clean machines when mnemonic or private-key exposure is possible.
Revoke durable trust artifacts after M365 or Entra compromise
For affected Microsoft 365 or Entra tenants and users, reset credentials and also revoke sessions and refresh tokens, remove unknown passkeys and MFA methods, audit OAuth and device-code grants, review app consents, constrain device-code authentication, require phishing-resistant authentication for high-risk users, and allowlist browser extensions.
Contain exposed ShareFile Storage Zone Controllers
Isolate or power off exposed or affected ShareFile Storage Zone Controllers now, reroute file-transfer workflows, and preserve logs or evidence where feasible.
Handle reported cloud and source-control secret exposures as control-plane credential incidents
Treat reported AWS, GitHub, GovCloud, Azure, and Accenture-style secret exposures as cloud control-plane credential incidents: rotate AWS/Azure keys and tokens, invalidate sessions, contain high-risk CI/CD and automation paths, review IAM privilege chains across S3/ECS/SQS/source control, quarantine exposed repositories and forks, and validate whether claimed leaked secrets were usable.
Freeze affected package lanes and audit build pipelines for reported malicious packages
Freeze only affected npm, NuGet, and package-build lanes; verify SBOMs, lockfiles, package names, versions, hashes, publishers, source repositories, and provenance; remove or replace reported affected packages after registry or vendor validation; rotate exposed CI, GitHub, cloud, payment, and wallet-related secrets if affected packages ran in developer or build environments.
Restrict Microsoft 365 trust-enrollment and session-persistence paths
Restrict new passkey, security-key, authenticator, phone, and recovery-method enrollment for high-risk users to managed devices, trusted networks, or verified help-desk workflows; limit broad device-code authentication where feasible; revoke suspicious sessions, refresh tokens, OAuth grants, and attacker-added authentication methods.
Use safety-led OT containment for Poland-style disruption
Operate in loss-of-view or loss-of-control mode: verify safe manual local operation first; freeze or tightly ACL remote and vendor access; preserve firewall, VPN, and HMI logs and images; surgically isolate enterprise, DMZ, remote-access, engineering, HMI, and controller networks; validate RTU and PLC firmware or controller logic and HMI project files; rotate default and vendor credentials; and reintroduce access only through logged jump hosts.
Microsoft 365 OAuth phishing requires token and consent revocation
Revoke Microsoft 365 user sessions and refresh tokens for exposed users, remove suspicious enterprise apps and service principals, revoke delegated OAuth grants, disable broad user consent, constrain device-code authorization, and require admin-reviewed consent for new OAuth permissions.
Raise targeted Android controls for high-risk users
For high-risk users, enforce managed Play allowlists, block unknown-source installs, monitor high-risk permissions such as Accessibility, notification access, default SMS, and battery-optimization exemptions, keep Firefox/WebView/Chrome and OS components current where managed updates are available, and focus controls on finance, executives, crypto handlers, journalists or NGO contacts, and India-facing teams.
Tighten developer supply-chain intake controls
Block unreviewed package intake through internal artifact proxies, require approval for new package names or maintainers, pin dependencies by digest, review lockfile diffs, harden GitHub Actions secrets, and verify raw Git objects and provenance rather than trusting platform labels or badges alone.
Constrain AI coding assistants for untrusted repositories
For untrusted repositories, disable or constrain assistant writes outside repository roots, disable auto-approval for file changes and command execution, block agent access to SSH keys, credential stores, cloud profiles, package-manager tokens, and CI/CD secrets, and use quarantine or ephemeral sandbox modes.
Isolate and hunt exposed Argo CD repo-server reachability
If the Argo CD repo-server gRPC path is reachable beyond the intended Argo namespace, move it into same-day isolation and hunting: enforce network-policy isolation, freeze or tightly control sync, and review repo-server logs, Redis access, odd manifests, and unexpected syncs.
Prioritize identity trust-state revocation over password resets
In the first 30 minutes of suspected identity exposure or token abuse, prioritize revoking sessions and refresh tokens, removing suspicious OAuth grants and service principals, restricting device-code auth, requiring phishing-resistant MFA for admins and high-risk users, and alerting on MFA, Conditional Access, OAuth, device, and impossible-travel anomalies.
Treat exposed Langflow and AI workflow services as credential-control-plane incidents
Take exposed Langflow or related AI/developer workflow services offline or restrict access while validating exposure, freeze risky automation, rotate reachable PostgreSQL, MinIO, config, cloud, CI/CD, API, model, and payment credentials, and restore only after fresh secrets and safer permissions are in place.
Authorize temporary emergency controls for identity, mobile, edge, and AI workflow exposure
Approve concrete 72-hour authority and downtime tolerance today: force identity controls for finance and admin users, allow security to quarantine non-compliant mobile devices without business-unit delay, and authorize emergency change windows for edge and AI workflow systems, with rollback and business-risk ownership.
Patch or compensate exposed ModSecurity multipart endpoints
Move exposed ModSecurity deployments to a fixed vendor or distribution build as soon as one is available, prioritize reported internet-facing i386 exposure for CVE-2026-52761, and treat multipart/form-data parser bypass as the broader exposed-app risk. Until patched, place reverse proxy or upstream WAF controls in front of upload and form endpoints, tighten or disable risky multipart routes, block malformed multipart requests, normalize line endings where supported, and run validation tests.
Run asset-share accounting incidents as first-24-hour containment
For Summer.fi/FleetCommander-style asset-share accounting incidents, prioritize first-24-hour containment over attribution debate: pause affected vault or strategy paths and related deposit, withdrawal, rebalance, mint, redeem, or share-conversion functions; snapshot balances and share supply; replay the exploit; review accounting, oracle, signer, upgrade, and treasury authority; notify users, exchanges, and bridges; and preserve traces before fixing forward.
Treat affected developer-tool executions as credential exposure
Treat developer workstations, CI runners, and build jobs that confirmedly installed or executed affected TeamPCP, ChocoPoCs, or malicious package artifacts as exposed. Freeze risky CI/CD execution paths, disable package install scripts where feasible, force installs through artifact proxies with lockfile or hash enforcement, rotate developer, cloud, source-control, package, and API credentials, and rebuild affected runners or machines from known-good images.
Sandbox agentic coding tools with shell, network, or file access
Organizations using coding agents with shell, network, or file access should run them in disposable sandboxes, disable unattended setup or script execution, block or alert on DNS TXT payload retrieval from developer workstations, and treat repository instructions in READMEs, issues, and package metadata as untrusted input.
Take same-day action for exposed REDCap in research and healthcare contexts
Research, healthcare, academic, medical, and military research organizations running legacy, unsupported, or internet-facing REDCap should take same-day action on the reported activity: restrict exposure, hunt for web shells and INFINITERED indicators, audit Google Workspace or mail routing/BCC/compliance rules, rotate REDCap-related credentials, and involve privacy/legal if sensitive patient or research data may have been accessed.
Treat OAuth, token, and infostealer exposure as attacker-held trust state
Treat ARToken, Klue OAuth, infostealer, and related credential exposures as attacker-held trust-state incidents rather than password-reset incidents. Revoke live sessions and refresh tokens, remove suspicious OAuth grants, kill remembered devices, rotate connected-app and integration secrets, revoke app passwords and personal access tokens, and review Microsoft 365, Salesforce, GitHub/GitLab, CI/CD, cloud, package, and mailbox activity for token use after supposed containment.
Govern agentic coding tools as untrusted code execution
Govern agentic coding tools as untrusted code execution: sandbox coding agents, block arbitrary network egress where possible, require human approval for shell and package operations, and keep secrets out of agent-accessible environments.
Showing 30 records · page 7 of 9
These controls filter only the 30 records on this page. Search the full Ledger →