Skip to content

Platform

IN DEVELOPMENT / DESIGN INTENT

A full-spectrum security operating system.

WARDOG is not an EDR wrapper or a generic SIEM. The engineering programme converges endpoint, identity, mail, network, cloud, SaaS, data, applications, attack surface and human-interaction risk on one governed substrate — with deterministic authority, explicit uncertainty and evidence for every consequential action.

Substrate
Graph · Brain · Gate · Response · Proof · Packs · Telemetry · Ledger
Authority
Gate is deterministic, deny-by-default and the sole privileged authority boundary.
Scope
Personal · SME · MSP · Enterprise

Programme disclosure

Capability without theatre.

This page is a public map of the canonical WARDOG engineering programme. It intentionally mixes evidence-backed foundations already exercised in the programme, active Stage 2 build-out and later-stage frontier direction. Roadmap and frontier items are not claims of live customer deployment, certification, production efficacy or completed provider integrations.

Evidence-backed foundation

Contracts, governance, simulation fixtures, tests, provenance rules and bounded response seams that have already been exercised in the engineering programme.

Active programme

Capabilities being implemented and hardened through the current staged queue. Provider adapters may remain simulated until explicitly authorised.

Frontier direction

Later-stage research and engineering targets that must remain measurable, reversible where applicable and bounded by Gate.

Shared substrate

One control plane for every security domain.

Modules can use replaceable open-source or commercial engines behind WARDOG-owned contracts, but evidence semantics, tenant boundaries and authority stay inside WARDOG.

Graph

A tenant-scoped security-state and interaction-risk graph linking assets, identities, messages, URLs, sessions, applications, workflows, evidence and requested actions.

Brain

Bounded analysis that can classify, explain, compare hypotheses and propose next steps, but cannot grant itself authority or infer human intent with certainty.

Gate

The deterministic, deny-by-default authority boundary for privileged response and controlled economic actions.

Response

Provider-neutral, reversible action proposals with bounded scope, expiry, replay protection and independent post-action verification.

Proof

Append-only evidence and provenance so observations, decisions, actions and verification can be reconstructed and challenged.

Pack Network

A governed path for signed detections, policies and response content to move from candidate research through testing, approval, shadow/canary and production.

Telemetry Fabric

A common ingestion and normalisation plane for endpoint, identity, mail, network, cloud, SaaS and other security observations.

Billing Ledger

Transparent action metering and economic authority, designed around attributable WT reservation, commit, release and compensating-credit events.

Coverage

Security modules designed to converge, not compete.

Every module feeds the shared graph, telemetry, proof and policy boundaries so WARDOG can reason across an incident instead of treating each tool as an isolated alert source.

Sentinel

Native AV/EDR/XDR, anti-ransomware, exploit-prevention and endpoint containment foundations.

Identity

Sign-in, session, token, device registration, privilege, application and directory defence.

Mail

Message authentication, phishing/BEC analysis, URL/QR/attachment evidence, mailbox hunt and bounded response proposals.

Network

Flow, DNS, HTTP/TLS, proxy, secure-web-gateway, Wi-Fi/captive-portal and traffic-manipulation defence.

Cloud

Cloud resource, identity, configuration, control-plane, workload and audit security.

SaaS

Cross-application posture, identity, configuration and activity controls.

Data / DSPM

Sensitive-data discovery, exposure context, movement and release-governance foundations.

Vulnerability

Asset-aware vulnerability, exposure, exploitability and remediation prioritisation.

Surface / EASM

External attack-surface discovery, ownership, exposure and change tracking.

Mobile

Mobile device, application, identity, network and risk-state coverage.

App / API

Application, API, dependency and software-supply-chain security.

SIEM

Normalised detection, correlation and evidence-backed security operations across modules.

Hunt

Bounded retrospective investigation across telemetry, graph context and provider-neutral search adapters.

Forensics

Evidence-preserving reconstruction, provenance and post-incident analysis.

Compliance

Control-state evidence, auditability and continuous proof rather than checkbox-only reporting.

Awareness

Human-centred security education informed by real attack patterns without turning users into the authority layer.

Supply Chain

Software, vendor and dependency risk with provenance and replaceable-engine boundaries.

Recovery

Verified restoration, rollback and recovery-state evidence after containment or disruption.

Personal

Consumer and family protection using the same governed substrate with appropriately reduced complexity.

MSP / Enterprise control

Tenant, customer, site, portfolio, budget and delegated-administration boundaries for managed and large estates.

Social Shield

Anti-social-engineering is a first-class security domain.

WARDOG's interaction-risk graph is designed to connect claimed identity, verified identity, message, URL, QR target, attachment, device/session, OAuth app, domain, voice/video persona, payment destination, requested action and organisational workflow. No model is permitted to infer human intent with certainty or grant authority.

Cross-channel threat coverage

  • Phishing, scams and business email compromise
  • Identity coercion and helpdesk social engineering
  • OAuth and device-code abuse
  • Adversary-in-the-middle token theft
  • QR, CAPTCHA and browser-in-browser phishing
  • AI-personalised spear phishing and AI-brand impersonation
  • Deepfake voice and video impersonation
  • Vendor and payment-change fraud
  • Captive-portal and traffic-manipulation lures
  • Government, diplomatic, defence, intelligence and research-targeted campaigns

Protected principals

  • Executives and board-level principals
  • Finance and payment approvers
  • Privileged administrators
  • Ministers, diplomats and government officials
  • Defence, intelligence and sensitive research personnel
  • Procurement, legal and HR
  • Other tenant-defined high-risk roles

High-consequence requests — credentials, device codes, OAuth scopes, payment/vendor changes, privileged resets and sensitive-data release — are designed to require stronger deterministic verification and out-of-band approval when policy demands it.

Emerging threat loop

Threat intelligence can propose. It cannot silently rewrite enforcement.

External research enters a governed lifecycle: source provenance → candidate TTP → map → simulate → false-positive test → approval → signed detection pack → shadow/canary → production. New reports never directly change enforcement.

2026 planning inputs

Current programme inputs include AI-enabled dynamic device-code phishing, adversary-in-the-middle token compromise, OAuth redirection abuse against government/public-sector targets, rapid QR phishing growth, CAPTCHA-gated phishing, AI-brand impersonation, state-linked captive-portal/traffic manipulation and nation-state AI-assisted reconnaissance/social engineering.

These are engineering and simulation inputs, not claims that every pattern is already covered in production.

Replaceable engines, owned authority

WARDOG can ingest capability from open-source, commercial and private engines behind owned contracts with licence, provenance, integrity, sandbox and shadow-evaluation controls. Engine tournaments can compare quality. No engine becomes the authority plane and no third-party report can bypass Gate.

Deterministic authority

The investigating system proposes. Gate disposes.

Privileged actions require evidence, policy authority, bounded scope, auditability, expiry/replay protection, reversibility where applicable and post-action verification. Missing evidence, malformed state or unavailable dependencies resolve to denial or explicit human approval — never permissive execution.

  1. 01

    Observe

    Continuous telemetry from endpoint, identity, SaaS, email and cloud.

  2. 02

    Reconstruct

    Events become a connected security-state graph, not a queue.

  3. 03

    Hypothesise

    Bounded reasoning proposes competing explanations.

  4. 04

    Verify

    Evidence is tested against thresholds before anything acts.

  5. 05

    Contain

    Only policy-authorised containment is permitted to execute.

  6. 06

    Remediate

    Root cause and persistence are removed, not just the symptom.

  7. 07

    Prove

    Post-remediation state is re-checked and recorded as evidence.

  8. 08

    Learn

    Verified behaviour improves detection without sharing raw data.

Transparent action billing

WT is an accounting unit, not a cryptocurrency.

WARDOG's working token unit is designed to make discretionary actions attributable and governable without turning safety into a paywall. Baseline fail-closed protection must remain available when discretionary WT balance is exhausted.

Attribution

Tenant → request/security context → action → Gate security/economic policy → execution → result → verification → tariff → ledger event.

Economic authority

Reservation, commit and release semantics with idempotency, anti-double-charge behaviour and compensating credits.

Budget controls

Spend ceilings, human-approval thresholds, MSP pooled budgets and enterprise cost centres without opaque surprise charging.

No premature payments

Real payment activation is deliberately deferred until ledger correctness and economic-authority behaviour are proven.

Frontier direction

Autonomous defence, bounded by measurable safety.

The later-stage programme pushes beyond alert handling toward continuous simulation, investigation, hardening and proof — while retaining deterministic authority and explicit uncertainty.

Continuous attack-path simulation
Adversarial multi-agent investigation
Counterfactual analysis
Autonomous hardening proposals and bounded execution
Digital-twin security simulation
Adaptive telemetry and retrospective hunting
Privacy-preserving collective defence
Deception orchestration
Breach forecasting with calibrated uncertainty
Blast-radius estimation and smallest-safe-change planning
Continuous proof of security state
Engine tournaments and shadow evaluation
Sovereign/private inference
Machine-speed containment for pre-authorised deterministic conditions

Frontier AI remains advisory or narrowly delegated under explicit deterministic policy. Machine-speed containment is reserved for pre-authorised conditions with measurable scope, evidence, verification and rollback semantics.

Request early access

WARDOG is in development. Tell us about your estate and your interest in early access.

Talk to us