Automotive Cybersecurity Engineering, from Architecture to Evidence.
KAVACH is Agnile’s AI-native ISO/SAE 21434 workspace for automotive cybersecurity engineering. It helps teams model vehicle architectures, identify cybersecurity assets, develop damage and threat scenarios, generate attack paths, map controls, and prepare traceable evidence with engineer-in-the-loop review. KAVACH supports ISO/SAE 21434 workflows beyond TARA while keeping engineers responsible for review, approval, and final evidence.
KAVACH does not replace engineering judgment. It structures and accelerates cybersecurity analysis while keeping engineers responsible for review, approval, and final evidence.
THE PROBLEM
TARA Is Still Too Disconnected from the Architecture.
In many programmes, cybersecurity analysis is performed across spreadsheets, documents, workshops, and disconnected evidence packs. Architecture changes are difficult to trace, threat coverage varies by analyst, and evidence often has to be reconstructed near review time.
KAVACH is built to make the cybersecurity workflow architecture-aware, traceable, reviewable, and easier to govern.
THE CHALLENGE
Where Manual TARA Hits Its Limits.
A modern vehicle carries 100+ ECUs, millions of lines of code, and dozens of network interfaces. Spreadsheet-based Threat Analysis can't keep pace with software-defined architectures — and UNECE R155 deadlines aren't waiting.
TARA cycles measured in weeks, not days
Threat enumeration, asset mapping, and risk scoring are still done by hand for most programs. The work is rigorous; the workflow doesn’t scale across platforms.
Lifecycle Evidence, Manual Traceability
ISO/SAE 21434 work-product evidence spans governance, concept, development, validation, production, operations, and TARA. When evidence lives across Word, Excel, and disconnected tools, traceability gaps often surface late during customer, assessor, or regulatory review.
Generic Threat Modeling tools weren’t built for vehicles
IT-grade tools don’t model ECU architectures, CAN/CAN-FD buses, automotive STRIDE, or ISO/SAE 21434 Clause 15 specifics. The gap shows up in Threat Scenarios you have to write yourself.
WORKFLOW
The KAVACH Workflow.
01
Model the architecture
Capture ECUs, interfaces, data flows, trust boundaries, communication paths, and relevant cybersecurity context.
02
Identify assets
Derive cybersecurity-relevant assets and properties from the modeled system context.
03
Develop damage scenarios
Connect compromised properties to stakeholder harms such as safety, operational, privacy, financial, and regulatory impact.
04
Generate threat scenarios
Identify plausible threat scenarios using automotive threat knowledge and architecture-specific context.
05
Build attack paths
Generate and review multi-step attack paths and attack trees that explain how threats may become feasible.
06
Map controls and evidence
Link risks to cybersecurity goals, requirements, controls, verification, and traceable cybersecurity evidence.
FEATURES
How KAVACH Stays Grounded.
Architecture-Aware TARA
TARA starts from the system model, not from a blank spreadsheet.
Curated Automotive Knowledge
Threat knowledge is grounded in automotive-specific sources, interfaces, ECUs, protocols, and attack patterns.
Engineer-in-the-Loop AI
AI assists analysis, but engineers review, edit, approve, and own the final outputs.
Traceability by Design
Assets, damage scenarios, threats, attack paths, controls, and evidence remain connected.
Deployment Flexibility
KAVACH supports desktop usage with AI inference options such as on-prem or customer-dedicated cloud VPC, depending on customer requirements.
LIFECYCLE COVERAGE
Beyond TARA: Lifecycle Evidence
Across Clauses 5 to 15.
Most approaches focus heavily on Clause 15 TARA. KAVACH supports ISO/SAE 21434 lifecycle evidence workflows across Clauses 5 to 15 — from organizational governance and project activities through concept, development, production, operations, post-development, and TARA.
Work-product labels are used as traceability anchors. Final evidence remains subject to engineering review, project tailoring, and customer approval.
Organizational Cybersecurity
Policy, governance, competence management, tool management, organizational audit
WP-05-01 – WP-05-05
Project Management
Cybersecurity Plan, Cybersecurity Case, Cybersecurity Assessment, release for post-development
WP-06-01 – WP-06-04
Distributed Activities
Supplier Cybersecurity Interface Agreement with responsibility allocation
WP-07-01
Continual Activities
Vulnerability Monitoring, cybersecurity event evaluation, managed vulnerabilities
WP-08-01 – WP-08-06
Concept Phase
Item Definition, Cybersecurity Goals with CAL, Cybersecurity Claims, Cybersecurity Concept
WP-09-01 – WP-09-07
Development & Validation
Cybersecurity Specifications, verification reports, integration testing, validation report
WP-10-01 – WP-11-01
Production
Production control plan with cybersecurity manufacturing requirements
WP-12-01
Post-Development
Incident response plan, end of cybersecurity support, decommissioning procedures
WP-13-01, WP-14-01
TARA Methodology
Full AI-powered TARA — Asset Identification through Risk Treatment with Attack Trees
WP-15-01 – WP-15-08
AI-POWERED TARA PIPELINE
Five Stages. One Generation Pass. ISO/SAE 21434 lifecycle coverage.
Each stage uses domain-specific AI grounding with citation provenance. Cross-stage analysis chains pass context from Assets through Risk Treatment without rebuilding it.
Assets
ECUs, data flows, interfaces with cybersecurity properties. SFOP impact ratings per Annex F.
Threats
AI maps curated automotive Threat Scenarios. STRIDE-classified with attack vectors.
Attack Paths
Attack Tree generation with feasibility per Annex G — time, expertise, knowledge, opportunity, equipment.
Risk Treatment
Risk Determination per Annex H. Controls mapped to each risk — encryption, authentication, monitoring.
Work-Product Evidence
Generate ISO/SAE 21434 work-product evidence — TARA Reports, Cybersecurity Case, Cybersecurity Goals — structured for audit review.
Assets
ECUs, data flows, interfaces with cybersecurity properties. SFOP impact ratings per Annex F.
Threats
AI maps curated automotive Threat Scenarios. STRIDE-classified with attack vectors.
Attack Paths
Attack Tree generation with feasibility per Annex G — time, expertise, knowledge, opportunity, equipment.
Risk Treatment
Risk Determination per Annex H. Controls mapped to each risk — encryption, authentication, monitoring.
Work-Product Evidence
Generate ISO/SAE 21434 work-product evidence — TARA Reports, Cybersecurity Case, Cybersecurity Goals — structured for audit review.
DEPLOYMENT
Two Deployment Models. Architecture Data Stays Inside Your Boundary.
KAVACH is a desktop platform with two options for AI inference. Pick the model that fits your security review and compliance posture.
OPTION 1
On-premise
Run KAVACH and the AI inference layer entirely on your own GPU infrastructure. No external network calls during TARA runs. Suited for teams with internal MLOps capacity and the strictest data-residency requirements.
- · Customer-provided GPU servers
- · Air-gappable; no inference data leaves the network
- · Customer manages the model lifecycle
OPTION 2
Customer-dedicated cloud VPC
A single-tenant VPC in the EU, provisioned per customer. Agnile runs the inference plane; you keep the data plane. Suited for teams that want managed infrastructure without sharing tenancy.
- · Single-tenant VPC, EU residency
- · Customer-owned encryption keys (BYOK)
- · Architecture models and project data stay within your tenant
In both models, architecture diagrams, threat scenarios, and Work Products remain inside the customer-defined boundary. The deployment option is a per-engagement decision, not a product SKU.
THE PLATFORM
See Every Screen. Every Work Product.
Watch KAVACH walk through the ISO/SAE 21434 TARA workflow — from item definition to evidence prepared for review.
THE SHIFT
From Spreadsheets That Age, to Evidence That Holds.
Manual TARA
Time to first TARA
4–8 weeks per system
Threat coverage
Bound to individual recall and spreadsheet limits
Risk Scoring
Reproducibility depends on who runs the analysis
Standard coverage
Clause 15 only (TARA work products)
Review readiness
Assembled at the end, often by hand
Repeatability across programs
Each program rebuilds artefacts from a blank template
KAVACH TARA
Time to first TARA
Hours
Threat coverage
AI retrieval over a curated automotive corpus
Risk Scoring
Structured per ISO/SAE 21434 Clause 15
Standard coverage
Clauses 5–15 (ISO/SAE 21434 lifecycle work products)
Review readiness
Built throughout, structured for audit review at any point
Repeatability across programs
Reusable templates, traceable lineage
See KAVACH on Your Architecture.
Bring a representative system, ECU, or feature. We’ll walk through how KAVACH moves from architecture to cybersecurity evidence.