PathVector — Attack Path Intelligence Engine by FrondeurLabs
Attack Path Intelligence Engine
Your attack surface,
mapped to every kill chain
PathVector ingests your ASM reports from FLS ORBIT and transforms them into complete, evidence-aware adversarial kill chains — predicting attacker techniques per phase, scoring candidate paths, and surfacing the route most likely to succeed against your environment.
Requires FLS ORBIT ASM — PathVector is the intelligence layer above your attack surface data.
Learn about FLS ORBIT →
7
Kill chain phases reconstructed end to end
3+
Ranked adversarial paths per ASM report
97%
F1-score on ATT&CK technique prediction (research)
160
Real-world threat reports validated in research
How it works
01
ASM Ingest
ORBIT ASM report feeds PathVector — assets, exposures, misconfigurations
02
Kill Chain Graph
ML ensemble predicts top ATT&CK techniques per kill chain phase with probability scores
03
Path Inference
MCTS + Policy-Value Network constructs ranked adversarial paths through your environment
04
Scored Output
Paths ranked by reward, cohesion, and transition plausibility — with payloads mapped
Kill Chain Graph — technique prediction
ASM Report → Top Predicted Techniques per Phase
PathVector reads your ORBIT ASM report and immediately predicts the most probable ATT&CK techniques across all seven kill chain phases — Recon through Actions on Objectives. Each technique is ranked by probability against your specific surface.
Kill Chain Visualization — full technique graph
The Kill Chain Graph renders every predicted technique as a node, connected across phases. Colour encodes phase; edge weight encodes transition probability. The visual makes the adversary’s decision space legible at a glance.
Kill Chain Inference — path construction & scoring
Transformer Baseline Paths
Three candidate kill chains, each scored. The transformer baseline selects a coherent technique per phase and ranks paths by semantic fidelity.
Symbolic Evaluation + Final Kill Chain
MCTS refines the baseline using a multi-objective reward — relevance, cohesion, transition plausibility. The final kill chain includes CVEs, CWEs, and exploits mapped per technique.
MCTS Thinking Search Tree
AlphaZero-style Policy-Value search over ATT&CK technique space
The MCTS search tree explores alternative adversarial interpretations of your environment before committing to a final path. Nodes are ATT&CK techniques; edges are phase transitions scored by the Policy-Value Network.
Payload mapping — Atomic Red Team
Per-technique payload resolution from Atomic Red Team
Every technique in the final kill chain is resolved to an Atomic Red Team payload — platform, atomic technique ID, and test command — so your red team can validate immediately without manual mapping.
What makes PathVector different
Surface-grounded path construction
Kill chains aren’t generic — they’re constructed from your ORBIT ASM data. The attacker’s starting point, techniques, and targets reflect what’s actually exposed in your environment.
MCTS + Transformer ensemble
Two independent reasoning engines — a transformer baseline and an AlphaZero-style MCTS — produce competing paths that are then reconciled by symbolic evaluation. No single model blind spots.
Evidence-aware scoring
Each path is scored across reward, relevance, cohesion, and transition plausibility. You see not just the winning path but why it scores higher than alternatives — and what CVEs make it viable.
Red team ready output
PathVector resolves every kill chain technique to an Atomic Red Team payload automatically. Your red team gets a test-ready playbook from a single ASM report — no manual technique-to-payload mapping.
Research foundation
KillChainGraph
Phase-aware ML ensemble (LightGBM + Transformer + BERT + GNN) — F1 97–99% across all phases
MDP-MCTS Inference
Policy-Value Network + Monte Carlo Tree Search over ATT&CK kill chain space — validated on FIN6, APT24, UNC1549
Atomic Red Team
Automatic payload resolution from technique ID to platform-specific test commands — no manual mapping
PathVector is the applied product built on two peer-reviewed FrondeurLabs research papers published on arXiv.
Read the research →
What you get from one ASM report
Technique probability map
Top ATT&CK techniques ranked by probability across all seven kill chain phases, grounded in your exposed surface.
Ranked adversarial paths
Three or more scored kill chains with symbolic evaluation metrics — reward, cohesion, transition plausibility, detections, mitigations.
Red team playbook
Every technique in the final kill chain resolved to an Atomic Red Team payload with platform, command, and test ID.
Who it’s built for
Red teams
Automated adversary simulation planning
Turn an ASM report into a prioritised, payload-mapped attack plan without manual technique research. PathVector does the kill chain construction; your team does the validation.
Threat intelligence
Surface-specific kill chain reconstruction
Move beyond generic TTPs. PathVector constructs kill chains that reflect what’s actually exploitable in your environment — grounded in ORBIT’s external and internal asset data.
SOC & detection
Know what to detect before the attack
PathVector surfaces the techniques most likely to appear in an attack on your environment. Prioritise detection rules around the paths that matter — not the whole ATT&CK matrix.
CISO
Evidence-based exposure narrative
A complete, scored adversarial path from your attack surface to your crown jewels — with CVEs, CWEs, and exploits — is a board-level risk story, not a posture statement.
We’re onboarding early design partners
If your team runs ORBIT ASM and wants to see kill chain inference on your own surface data, we’d like to talk.
info@frondeurlabs.com
Requires FLS ORBIT ASM · Pre-launch · Not for distribution