The Cyber Toolchain — issue 123, October 28, 2026

Adds -sV / -service-version flag for service version detection using an embedded nmap-service-probes database, with -sV-fast to limit probing to port-hinted services only, -sV-timeout to control probe timeout, -sV-workers to set concurrency, and -sV-probes to supply a custom nmap-service-probes file path.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

$ naabu -host example.com -top-ports 1000 -sV -sV-workers 10 -sV-timeout 10s -json -o results.json

Adds -uP / -udp-probes flag to send protocol-specific payloads on UDP scans, using the nmap-service-probes database to identify UDP services.

$ naabu -host 192.168.1.0/24 -p 53,161,500,4500 -s SYN -uP -json

Adds -show-dead flag to report hosts that did not respond during host discovery; requires host discovery to be enabled to surface asset inventory gaps.

$ naabu -host 10.0.0.0/24 -sn -show-dead -silent

Endor Labs introduced AURI Fix for Open Source, a private-beta remediation system triggered by the new endorctl fix CLI command, which automatically assesses a reachable finding and opens a pull request with the safest fix — an upgrade, a first-party code mitigation, a targeted dependency patch, a configuration change, or dependency replacement.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

$ endorctl fix

Renames Suspicious Files to Malicious Files and adds a Rules page under Developer Machines listing every detection rule with its campaign, indicator groups, and file patterns, searchable by rule name and filterable by campaign, with per-rule activate/deactivate controls that stop scanning across all enrolled devices and move existing findings to a reversible Suppressed state.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠ A new Findings page under Developer Machines groups flagged files into New, Resolved, and Suppressed tabs with device, user, OS, campaign, confidence, and status metadata, filterable by confidence or campaign and exportable to CSV.

⤢ ⤢ ⤢

Dev Machine Guard gains a plugin detail panel listing marketplace source, plugin file origin, per-device installation scope (User, Project, Local, System), version, state, and component breakdown by kind (skills, commands, MCPMCPModel Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. servers), with Claude Code skill usage counts; plugin-sourced skills and MCP servers are attributed back to their originating plugin on the existing Agent Skills and MCP Servers pages. A new Agent Plugins page under GitHub > AI Agents inventories plugin manifests (.claude-plugin/plugin.json for Claude Code, .codex-plugin/plugin.json for Codex) published across org repositories, showing publishing repository, plugin folder path, declared metadata, branch, and commit scanned, plus a Used by view listing every repository whose settings enable the plugin for all contributors. Supports CSV export of the Dev Machine Guard inventory, and filtering by installation scope/source kind/enablement (Dev Machine Guard) or by agent/repository (GitHub).

⤢ ⤢ ⤢ ⤢

Adds a Malicious Files section under GitHub scans that inspects full repository history — every branch, pull request, and tag — for files planted by known supply chain attack campaigns such as Shai-Hulud, GlassWorm, and Miasma. Includes a Rules view showing each rule's campaign, severity, revision, and ordered checks; a Scan Status view showing the most recent scan per repository with commits/file versions scanned and last/next scan times; and a Findings view listing matched files with branches, pull requests, tags, matched checks, and the introducing commit, filterable by severity, campaign, or rule, with CSV export.

The Threat Intel → Campaigns page now shows a per-organization status badge on every malware campaign — Blocked by Socket, Detected in Firewall/SCA, or Not Detected — computed from repository scans and Firewall event traffic.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠ A new Org Impact tab on each campaign lists each affected package and version separately along with the repositories, branches, and first/last-seen dates, which link directly to the originating scan report or matching Firewall event; new Blog and Affected Packages tabs add Socket write-ups and the full list of affected packages to the campaign detail view.

⤢ ⤢ ⤢

The Java code graph now reads anonymous classes, local classes, and enum constant bodies as first-class definitions, with scoped local-class names and qualified this binding.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

GitHub Apps installed on organizations can now authenticate to the Copilot SDK using a short-lived installation access tokenaccess tokenA credential that authorizes a client to use a service or API. It usually has limited permissions and a lifetime; anyone holding it may be able to use those permissions. instead of a user's credentials (public preview), enabling agents and automated services to call Copilot under an org identity.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠ A new read-only Copilot Requests organization permission scopes what an app can do when authenticating as this workload identityworkload identityAn identity assigned to software, such as a service or container, rather than a person. It lets the workload authenticate and receive permissions without sharing a user's credentials., and app-driven usage is attributed to the installing organization — including model policies, budgets, and spending controls — with app traffic identifiable in AI usage CSV reports via the app_client_id field.

Veracode CLI 2.53.0 adds veracode scan static for running static scans and reviewing security findings within a Git workflow, veracode findings static for accessing cached static scan findings in the local Veracode directory, veracode sca for SBOMSBOMA software bill of materials, an inventory of the components and versions inside an application. It helps teams find affected software when a dependency has a vulnerability or licensing issue. generation via SCA agent-based scans publishing results in CycloneDX 1.6CycloneDX 1.6An OWASP-maintained Software Bill of Materials standard that catalogs components, dependencies, and supply-chain metadata; tools use it to produce or consume machine-readable SBOMs in a widely supported schema. and SPDX 2.3SPDX 2.3An open standard from the Linux Foundation for describing software bills of materials, letting cyber tools exchange dependency, license, and vulnerability data in a common machine-readable format. formats, and auth login for onboarding new users via OAuth2OAuth2An open authorization framework (RFC 6749) that lets a service grant limited access to a user's resources on another service without exposing credentials, giving cyber tools a standard way to authenticate against third-party APIs..⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

$ veracode scan static && veracode findings static

veracode package adds the --csharp-source flag to package C# source code directly for Static Analysis without compiling it into binaries, the --skip-packagers flag to exclude specific artifacts (such as unsupported SQL) from packaging without affecting other languages, the --exclude flag to prevent specified Java packages or JAR files from being scanned, the VERACODE_PACKAGE_GOLANG_GENERATE environment variable to enable client-side CGO code generation when packaging Go applications, the SRCLR_MAKE_WINDOWS environment variable to preprocess C/C++ source files with Make on Windows using a native Windows gcc or clang compiler, the --java-source flag and VERACODE_PACKAGE_JAVA_SOURCE=true environment variable to enable Java source code packaging, support for requirements-dev.txt in Python packaging workflows, support for the Poetry package management tool for Python autopackaging, automatic inclusion of .yaml, .yml, and .properties files when packaging Java source code, support for packaging Go v1.24, and support for a generic packager for Apex.

Veracode deprecates .NET cleansers for flaw classes CWECWECommon Weakness Enumeration, a catalog of recurring software and hardware weakness types. It describes classes of mistakes, such as injection, rather than individual vulnerability instances.-93, CWE-113, and CWE-117 across antixsslibrary.dll: Microsoft.Security.Application.AntiXss.HtmlAttributeEncode, antixsslibrary.dll: Microsoft.Security.Application.AntiXssEncoder.HtmlAttributeEncode, antixsslibrary.dll: Microsoft.Security.Application.Encoder.HtmlAttributeEncode, antixsslibrary.dll: Microsoft.Security.Application.Encoder.HtmlEncode, mscorlib.dll: System.Security.SecurityElement.Escape, system.dll: System.Net.WebUtility.HtmlEncode, system.web.dll: System.Web.HttpServerUtility.HtmlEncode, system.web.dll: System.Web.Security.AntiXss.AntiXssEncoder.HtmlEncode, system.web.dll: System.Web.Util.HttpEncoder.HtmlAttributeEncode, system.web.dll: System.Web.Util.HttpEncoder.HtmlEncode, system.web.mvc.dll: System.Web.Mvc.HtmlHelper.AttributeEncode, system.web.mvc.dll: System.Web.Mvc.HtmlHelper.Encode, system.windows.browser.dll: System.Windows.Browser.HttpUtility.HtmlEncode, system.windows.dll: System.Net.HttpUtility.HtmlEncode, and System.Runtime.dll: System.Net.WebUtility.HtmlEncode; applications relying on these as cleansers will no longer have those flaws suppressed.

The Java API Wrapper adds support for the -includenewmodules parameter and enables sending requests to the Identity REST APIREST APIA web API design style where clients interact with resources over HTTP using standard methods like GET, POST, and DELETE, giving cyber tools a predictable, language-agnostic way to integrate with external services. from within the wrapper, while Veracode C# API Wrapper 23.5.8.8 adds proxy support via the https_proxy environment variable, accepting host, port, and credentials.

$ export https_proxy=http://proxyuser:proxypass@proxy.corp.example.com:8080

veracode package discover detects toolchains required to create scannable artifacts, including JavaScript projects using npm or Yarn and Java projects using the Gradle package manager, and introduces exit codes 5, 6, and 7 to improve CI/CD pipeline automation and error handling.

veracode scan now categorizes secret-type findings as Critical, prioritizes vulnerability results from the NVDNVDThe National Vulnerability Database, maintained by NIST, is the U.S. government repository of CVE records enriched with CVSS scores and metadata; cyber tools query it to drive vulnerability prioritization. database when available, enables applying a custom policy defined in the Veracode Platform and viewing policy evaluation for individual findings, enables specifying a custom location from which vulnerability databases can be downloaded, adds support for parsing Dockerfile heredoc syntaxes, and introduces schema enhancements for container and IaCIaCInfrastructure as Code: the practice of defining and provisioning compute infrastructure through machine-readable config files rather than manual processes, giving cyber tools repeatable, version-controlled environment setup. scan results.

Pipeline Scan adds an --include parameter to specify top-level modules to include during scanning, with prescan module discovery now shown in results; available in Veracode CLI version 23.4.3-0 and Docker image version 23.4.3.

The Veracode Azure DevOps Extension's Upload and Scan task gains two optional parameters: scanpollinginterval to control how often the task polls for scan status, and lifecyclestage (not supported in YAML pipelines).

- task: Veracode@3
  inputs:
    ConnectionDetailsSelection: 'Endpoint'
    AnalysisService: 'VeracodeEndpoint'
    veracodeAppProfile: 'MyApp'
    version: '$(build.buildNumber)'
    filepath: 'drop/myapp.zip'
    scanpollinginterval: '60'

veracode fix sca supports asynchronous Fire-and-Forget (FnF) mode, submitting remediation requests for background processing without waiting for fixes or pull requests.

Python applications with flaws previously reported as CWE-201 that should have been CWE-918 are being recategorized, potentially breaking existing flaw matching and requiring new mitigations.

The Veracode Jenkins Plugin adds a Show Unstable Status for Failed Policy Evaluation option that marks a job as Unstable when a scan completes but fails the security policy.

A new DAST Scanner integration appears under Integrations > Scanners > WEB, letting teams schedule authenticated web scans and manage targets from within Phoenix.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠ Each target is configured with a target URL, start time, time zone and schedule, with multiple targets added via + Add New Target, and authentication via the Auth Token (Bearer) field supports bearer and JWTJWTAn open standard (RFC 7519) for encoding claims as a signed, optionally encrypted JSON object passed between parties. Cyber tools use it to verify identity and permissions without querying a central session store. tokens (cookie, basic, and client-certificate auth are not supported). The integration page monitors target status — Active/Fetched, In Progress, Not Started, Disabled, Missing, and Error — alongside last fetch date, and exposes Fetch and Fetch/pause all controls to enable or pause individual or all targets, plus an optional setting to auto-delete missing targets not reported for a configurable number of days.

⤢ ⤢

Signal Agentic Fix is an AI agent that performs end-to-end vulnerability remediation—scan, triage, fix generation, validation, and pull-request submission—without human bottlenecks.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠ It reuses the same deterministic SASTSASTStatic application security testing, analyzing source code or compiled code for vulnerabilities without running the application. It helps developers find problems before deployment. and SCASCASoftware composition analysis, identifying third-party dependencies and checking them for known vulnerabilities and license issues. It covers software a team uses rather than writes itself. engines from Polaris, Coverity, and Black Duck SCA for consistent scan results inside the agent workflow, and is available in early access as a Claude Code plug-in for Coverity users, with additional coding-agent and DevOps pipeline support planned for Coverity, Polaris, and Black Duck SCA. A planned addition, Signal AI Code Analysis support, will enable deeper discovery of novel risks such as IDORIDORInsecure Direct Object Reference, a web vulnerability where an application exposes internal identifiers (IDs, filenames) without access checks, letting an attacker substitute another user's ID to reach their data. and BOLABOLABroken Object Level Authorization, an API vulnerability class where a server fails to verify that the requesting user owns or may access the object ID they supply, letting attackers read or modify other users' data. within agent runs.

The POST /api/v4/ai/duo_workflows/workflows endpoint (generally available from GitLab 19.5) lets any external system trigger a custom Duo Agent Platform flow over HTTP by passing project_id, ai_catalog_item_consumer_id, and a goal.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠ Request body fields allow_agent_to_request_user, agent_privileges, and pre_approved_agent_privileges control whether the flow executes autonomously and which tool permissions it is granted. The response includes a web_url field linking to the agent session for inspecting each step the flow took, and the flow's ai_catalog_item_consumer_id can be looked up via a GraphQL query.

⤢

$ curl --request POST https://gitlab.com/api/v4/ai/duo_workflows/workflows \
  --header 'PRIVATE-TOKEN: <your_project_access_token>' \
  --header 'Content-Type: application/json' \
  --data '{
    "project_id": 12345,
    "ai_catalog_item_consumer_id": 67,
    "goal": "Class: OpportunityDiscountService\nException: System.NullPointerException\nMessage: Attempt to de-reference a null object\nStack trace: ...\n",
    "start_workflow": true,
    "allow_agent_to_request_user": false,
    "agent_privileges": [1, 2, 3],
    "pre_approved_agent_privileges": [1, 2, 3]
  }'

Custom agent flows are defined as a single YAML file using flow registry v1 syntax, specifying AgentComponent agents, toolsets (find_files, grep, read_file, list_commits, create_issue), and prompt templates that drive the agent's behavior.

version: "v1"
environment:
  ambient
components:
  - name: "sf_error_triage"
    type: AgentComponent
    prompt_id: "sf_error_triage_prompt"
    inputs:
      - from: "context:project_id"
        as: "project_id"
      - from: "context:goal"
        as: "error_report"
    toolset:
      - "find_files"
      - "grep"
      - "read_file"
      - "list_commits"
      - "create_issue"
prompts:
  - prompt_id: "sf_error_triage_prompt"
    prompt_template:
      system: |
        You triage errors for a team. Find the source file, read the code,
        list recent commits, and create one issue with root-cause analysis.
      user: |
        Project ID: {{project_id}}
        Error report: {{error_report}}
      placeholder: history
routers:
  - from: "sf_error_triage"
    to: "end"
flow:
  entry_point: "sf_error_triage"

SonarQube Cloud's Issues page now supports the hunter-agent tag and category tags like hunter-broken-access-control to filter Hunter Agent findings (broken access control, business logic flaws, auth gaps) alongside the existing Fixable by Agent filter; up to 20 findings can be selected and sent to Fix with Agent.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠ Each Hunter Agent finding generates its own individual pull request (or merge request on GitHub), unlike the grouped-by-rule-key behavior used for rule-based issues. For GitHub ActionsGitHub ActionsGitHub's built-in CI/CD system that runs automated workflows triggered by repository events, letting cyber tools automate testing, building, and releasing without managing separate pipeline infrastructure. pipelines, CI follow-up is enabled on these fix PRs: when CI fails, the agent reads build logs, pushes a corrective commit, and retries up to three times before leaving the PR open. This capability is also brought to SonarQube Server 2026.5, with both the Hunter Agent and Remediation Agent able to run on self-deployed infrastructure.

Adds --aws-region to the scan and standalone toon commands to expand AWS discovery beyond the SDK-selected region, accepting up to eight explicit regions (global IAMIAMIdentity and Access Management — a framework of policies and services for controlling who can authenticate to systems and what resources they are authorized to use, letting cyber tools enforce least-privilege and audit access./S3 resources are inventoried once).⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠ Adds --gcp-project to scan and toon to expand GCP discovery beyond the home project, accepting up to seven additional projects, each using its own visible IAM context.

$ kingfisher scan --aws-region us-east-1 --aws-region eu-west-1 --aws-region ap-southeast-1 ./key.json

$ kingfisher toon ./key.json --gcp-project production-project --gcp-project staging-project

Blast-radius output now includes automatic risk and coverage assessment in scan and toon (no extra flag needed), labeling permissions by capability category — e.g. s3:ListBucket as Metadata read, s3:GetObject as Sensitive data read — and distinguishing direct access, candidate role paths, and observed API access from mere policy grants or conditions. This is exposed via the new provider_metadata.authorization_evidence.assessment field, containing schema version, assessment timestamp, capability categories, evidence labels, coverage, resource explanations, recommendations, and notes; capability_risk is tracked separately from the original severity field. Detailed assessments are capped at 4096 capability records and 1000 resources per identity, with up to 256 permissions per resource, and legacy report data is retained separately. The observed_at timestamp is intentionally excluded from snapshot comparisons since it varies between runs.

Adds a blast-radius HTML investigation viewer with Overview, Findings, Coverage, and Blast Radius views that opens existing JSON/JSONLJSONLA text format where each line is a self-contained, valid JSON object, making it easy to stream, append, and process large datasets line by line without parsing a full JSON array into memory. reports and supports drag-and-drop via a compact import bar. It groups identities by provider and surfaces policy provenanceprovenanceEvidence recording where a software artifact came from and how it was built, including its source and build process. It helps a consumer check that the artifact matches a trusted build., candidate paths, conditions, exclusions, and observed API methods alongside a separate resource map. A synthetic investigation report illustrating AWS, GCP, Azure, Hugging Face, and GitHub evidence can be opened with kingfisher view or the report picker to explore the viewer before running a real scan.

$ kingfisher view

Adds a --background flag to ggshield machine scan and ggshield machine report to run scans at low priority using 2–4 threads and one-tenth of normal memory (minimum 512 MB), suited for scheduled MDMMDMMobile device management, centrally configuring and managing enrolled devices. Administrators use it to apply settings, distribute applications and enforce device security policies. jobs.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠ Adds scan.resources.background.mode: auto to machine_scan.yaml to make background-priority scanning the default for every scan regardless of how it's triggered.

$ ggshield machine report --background

scan:
  resources:
    background:
      mode: auto

ggshield machine dashboard gains an Overview page that lists one row per cleanup area (AI agents, repositories, history and logs) and guides remediation order: overprivileged secrets first, then history leaks, then secrets in daily use.

$ ggshield machine dashboard

The new --max-response-size MiB flag stops reading any HTTP response once it exceeds the specified size, guarding against memory exhaustion on targets serving huge or endless responses; default is 16 MiB, max is 2047, and 0 disables the cap entirely.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠ This changes default behavior: responses larger than 16 MiB are now truncated by default, so existing scans relying on unbounded reads must pass --max-response-size 0 to restore the previous behavior. The option can also be persisted via max_response_size under cli_options in a project config file such as pwd/.wpscan/scan.yml.

$ wpscan --url https://target.tld --max-response-size 64

cli_options:
  url: 'https://target.tld'
  max_response_size: 64

Adds Settings > AI > Providers to configure OpenAI, Anthropic, Google Gemini, SpaceXAI, OpenRouterOpenRouterA unified API gateway that routes requests to dozens of LLM providers under one endpoint, letting cyber tools swap or compare models without rewriting provider-specific integration code., and Amazon Bedrock from one place with credential testing before use.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠ Also supports custom providers (e.g. a local LM Studio endpoint) by setting a Provider ID, matching API type, Base URL, and an optional API Key, with a Test action to verify the connection.

⤢

Adds an Intercept > Capture menu (marked Experimental) with three recording modes — All traffic, In-scope only, and Custom rules — plus a Passthrough list under Custom rules mode so hosts on it pass through Caido without being recorded. In-scope only uses the active Scope definition to silently skip recording out-of-scope requests.

⤢

The Fraud SDK .Net 2.1.0 adds an optional eventDate field to the event payload to specify the effective event time, useful for replaying or backfilling events rather than relying on submission time.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

{
  "eventDate": "2026-10-27T14:32:00Z",
  "...": "..."
}

Adds Request::get_vpn_proxy_data, Request::get_vpn_proxy_*, and RequestHandle::vpn_proxy_* methods to expose metadata for detecting whether a downstream client is residential, a VPNVPNA virtual private network, an encrypted tunnel between devices or networks. It carries traffic across another network as though the endpoints had a private connection., or a proxy.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

When Request::with_image_optimizer is used on a service where Image Optimizer is not enabled, the origin response is now passed through unmodified with a Fastly-Io-Error response header instead of failing.

The -Z, --timezone <OFFSET> flag was added to dfir-timeline, search, eid-metrics, logon-summary, log-metrics, and extract-base64 to render output timestamps with a fixed UTC offset in signed +HH:MM or -HH:MM form (e.g.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠ --timezone "+09:00" or -Z "-01:00"). It conflicts with --utc and --iso-8601, and filtering, sorting, and histogram bins remain UTC-based.

Adds a POST /api/public/v0/notifications/subscriptions endpoint to subscribe to email notifications for every scan in a project programmatically.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠ The user who started a scan is automatically emailed on completion or failure, and users can subscribe at three scopes — a single scan, every scan in a Folder, or every scan in a Project — including a Notify me option on a running scan to subscribe mid-flight. A new Notification Settings view lets users choose which events to receive and filter subscriptions by scope, while organization administrators can enable or disable the scan email channel org-wide under Integrations with a single switch that reflects delivery status even after a recipient has used an unsubscribe link.

Adds PATCH /v0/licenses and DELETE /v0/licenses (with id as a query parameter) to edit, deprecate, restore, or delete a custom license programmatically. The same actions are available in the Settings > Policy > Licenses UI, with audit trail entries recorded for create, update, and delete.

$ curl -X DELETE 'https://platform.finitestate.io/v0/licenses?id=<license-id>' \
  -H 'Authorization: Bearer <token>'

Extends GET /api/public/v0/components and the component detail endpoint to return the exclusion comment, the time of exclusion, and the author who excluded the component. Each exclude or include action — including the comment and full component count — is recorded in GET /v0/audit, and the comment, excluding user, and timestamp are surfaced on the Component view and the Excluded list in the UI.

$ curl 'https://platform.finitestate.io/api/public/v0/components/<component-id>' \
  -H 'Authorization: Bearer <token>'

$ curl 'https://platform.finitestate.io/v0/audit' \
  -H 'Authorization: Bearer <token>'

Rebuilds the Email Identity Protection report to place every message into exactly one of three mutually exclusive outcomes — delivered without verification, held before delivery, or blocked — with a secondary breakdown of held-message outcomes (retrieved, reported unrequested, failed verification, never attempted, verification outstanding).⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠ Separates activity by message type (password resets, verification codes, account activations, account access messages), includes a service matrix showing each service's setting (Protect, Block, Detect Only, or Disabled), and adds PDF export for leadership reviews.

Adds the Auto-classify OAuth apps setting (off by default) that automatically applies the OAuth Agent's recommended classification to new and audited app connections and triggers configured remediations without manual review. Adds a Classification State filter and column (values: Suggested or Applied) to the OAuth apps list and CSV exports, plus a new OAuth Agent activity report summarizing how apps have been classified, which have been remediated, and which require further investigation.

Adds a Blast radius card on each OAuth app's detail view showing how much sensitive content (files and messages) the app's connected accounts can reach, with links from either count into the Explorer view to inspect the specific at-risk content.

Adds Vendor Trust Level and Scope Risk Level columns to the OAuth Security Issues view for sorting and filtering apps by reputation and permission breadth.

Replaces the previous Email Redaction and Sensitive Content reports with a unified Email Redaction report that includes an estimate of currently redacted messages with a weekly trend, histograms of access requests and retrieved-message age, and a chart of access requests over time.

Attack Score Flash now runs on all inbound mail, producing high-precision Malicious, Suspicious, Spam, and Graymail verdicts to catch messages existing rules may miss, similar to Attack Score.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠ Potentially Malicious and Spam detections are promoted into the core feed so custom rules are no longer needed to surface them. Existing automations that trigger on Attack Score verdicts now also act on messages flagged by Attack Score Flash, increasing automated flagged-message volume, and the Autonomous Security Analyst automatically reviews Attack Score Flash Suspicious verdicts via the existing Send suspicious messages to ASA automation. Potentially Malicious detection is Sublime Enterprise-only, while Attack Score Flash itself is available on both Sublime-managed cloud and self-hosted deployments.

A free read-only Containment Assessment at aviatrix.ai/containment-assessment installs nothing and can reserve AWS Marketplace enrollment.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠ Clearance is free for all enterprises — covering Outbound Containment, Lateral Containment, and Intrusion Prevention with no workload cap — from November 1, 2026 through May 1, 2027 via AWS Marketplace. A General Availability tier offers the Containment Assessment plus 15 workloads in monitor or enforce mode with no time limit, priced per workload and delivered on Aviatrix Cloud (same controls also available on Aviatrix Enterprise). Aviatrix Active Defense is available today in free early access for design partners.

Aviatrix Clearance™, part of the Aviatrix Containment Platform, introduces Outbound Containment (limits each workload to outbound destinations it is explicitly cleared for, named by domain rather than network address) and Lateral Containment (restricts workload-to-workload reachability by identity across clouds and Kubernetes/serverless paths not governed by hub-based inspection). It enforces per-workload outbound and lateral communication policy by workload identityworkload identityAn identity assigned to software, such as a service or container, rather than a person. It lets the workload authenticate and receive permissions without sharing a user's credentials. and protocol across clouds, down to individual containers, serverless functions, databases, and AI agents.

▶ ⤢

Adds GetDecisionsSinceCountByScenario expression helper for querying per-scenario decision counts, usable in postoverflow or rule expressions to gate actions on how many decisions a scenario has already produced in a time window, e.g.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠ GetDecisionsSinceCountByScenario('crowdsecurity/http-probing', '1h') > 5, and LookupHost DNS cache support in expression helpers to reduce repeated DNS lookups.

$ GetDecisionsSinceCountByScenario('crowdsecurity/http-probing', '1h') > 5

Adds SkipProcessing() to the WAF for conditionally bypassing rule evaluation, letting a matching request short-circuit all further rule evaluation — useful for allowlisting trusted internal scanners.

$ SkipProcessing()

Adds a cscli bot subcommand for managing bot detection from the CLI, a FlagSubmission() helper for marking submission events, custom JS fingerprinting allowing user-defined and hub-distributed detection rules, and optional SameSite=None and Partitioned flag support on bot detection cookies.

$ cscli bot

Adds a cs_features metric to track which Security Engine features are active, a kind label on the cs_alerts metric, and a CAPI last-pull timestamp plus error counter to metrics.

Adds PLERION-CLOUDTRAIL-84 (Critical) to detect compromised AWS keys via AWS Health exposed/compromised access key reports, naming the key and IAMIAMIdentity and Access Management — a framework of policies and services for controlling who can authenticate to systems and what resources they are authorized to use, letting cyber tools enforce least-privilege and audit access. user (requires us-east-1 in your CDR regions); PLERION-CLOUDTRAIL-85 (High) to detect deletion of Amazon Bedrock model invocation logging, a LLMjacking cover-track technique; and PLERION-AWS-1252 (Low) to check whether CloudTrailCloudTrailAWS service that records API calls and account activity across an AWS environment, giving cyber tools an audit log to detect anomalous or malicious actions. logs data events for AWS::BedrockMantle::Project, the endpoint for OpenAI-compatible and Anthropic Messages APIs, via an advanced event selector on a multi-region trail.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

Adds PLERION-AZURE-463 (Medium) to detect Azure Function App runtimes past Microsoft's end-of-support date, covering .NET, .NET isolated, Java, Node.js, PowerShell and Python; flags Python 3.8/3.9, .NET 6, PowerShell 7.2 and Node.js 20 on next scan.

Expands the Access logging filter on the Assets page and the matching custom-report field to cover API Gateway (PLERION-AWS-494, PLERION-AWS-799, PLERION-AWS-801), load balancers (PLERION-AWS-791, PLERION-AWS-792, PLERION-AWS-795), VPCs (PLERION-AWS-37) and EKS clusters (PLERION-AWS-107).

Widens PLERION-AWS-231 CloudFront general-logging check to pass distributions whose standard logging (v2) delivers via a CloudWatch Logs delivery to CloudWatch Logs, Amazon S3 or Amazon Data Firehose, and widens PLERION-AZURE-74 Key Vault logging check to pass vaults whose diagnostic setting enables the audit or allLogs category group, not only an explicit AuditEvent selection.

aya-ebpf-macros v0.4.0 adds support for BTF map pinning by name in aya-ebpf, letting maps be pinned to the BPFBPFA Linux kernel virtual machine that lets programs attach small, sandboxed bytecode filters to kernel hooks — syscalls, network events, tracepoints — so cyber tools can observe or redirect kernel behaviour without writing a kernel module. filesystem using their BTF-derived names, and aya now validates pinned maps before reuse by checking type and size correctness when attaching to an existing pinned map.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

Adds lifecycle.idlePolicy create field with idleTimeoutSeconds (30–86400), onIdle: pause, wakeOnAccess, and optional activeProbe to enable auto-pause and wake-on-access for fast sandboxes; server-side config moves under [automation.idle_pause], requiring redis.dsn when the sweeper is enabled, and the idle sweeper now also supports an OpenSandbox-Skip-Activity sentinel header to suppress one activity observation, stripped on both gateway and proxy hops.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠ All five SDKs gain typed client support with fail-fast validation and a wake-retry contract: on a 503 with OpenSandbox-Wake: expired, SDKs retry after Retry-After up to a bounded budget (default 5 attempts / 60s) and raise a typed wake-timeout exception on exhaustion, with stream-start wake-retry active in JavaScript, Python, and Go.

$ curl -X POST https://api.opensandbox.example/v1/sandboxes \
  -H 'Content-Type: application/json' \
  -d '{
    "template": "my-template",
    "lifecycle": {
      "idlePolicy": {
        "idleTimeoutSeconds": 300,
        "onIdle": "pause",
        "wakeOnAccess": true
      }
    }
  }'

[kubernetes]
template_mitm_ca_secret = "mitm-ca-secret"

[automation.idle_pause]
enabled = true

[redis]
dsn = "redis://localhost:6379/0"

Adds [kubernetes] template_mitm_ca_secret server config (default off) to bake a pinned MITM CA into server-created SandboxTemplates via spec.egressMitmCertificate, so an egress sidecar restart no longer rotates the trust anchor under live sandboxes.

[kubernetes]
template_mitm_ca_secret = "mitm-ca-secret"

[automation.idle_pause]
enabled = true

[redis]
dsn = "redis://localhost:6379/0"