Join our Newsletter — 33% off our NHI Course
Home› Guides› Automotive TARA Guide: A Worked Threat Analysis for…
Guide OT & IoT Security

Automotive TARA Guide: A Worked Threat Analysis for a Telematics ECU

PDF for newsletter subscribers · free to read online

← All guides
By NHI Mgmt Group, edited by Lalit Choda Updated 14 November 2026 17 min read
On this page

Most automotive threat analysis and risk assessments (TARAs) fail not on the method but on its application: assets are listed as functions, feasibility scores are guessed, and the treatment decisions cannot be traced back to anything an assessor can check. This guide walks through a complete TARA for one example electronic control unit (ECU), a telematics control unit (TCU) with cellular connectivity and over-the-air (OTA) update support, from asset identification to cybersecurity goals. It shows the working for every rating and treatment decision. It assumes you already know where the TARA sits in ISO/SAE 21434 and UN Regulation No. 155; for that context, read the ISO/SAE 21434 Automotive Cybersecurity Engineering Guide first. It is written for product security engineers, system architects, supplier cybersecurity managers and assessors who review TARA work products.

Key takeaways

  • Scope the item before you list a single threat. A TARA is only as good as its item definition: boundaries, interfaces, operational environment and the external systems the ECU talks to.
  • Treat credentials as assets in their own right. Keys, certificates, seed and key secrets and back-end service account credentials carry the highest-impact damage scenarios in most connected ECUs.
  • Score feasibility factor by factor, and write the reason down. An attack potential score without a justification per factor cannot be reviewed, repeated or updated when the threat landscape changes.
  • Calibrate the method for scale. Attack potential scoring can rate a hard, remote, fleet-wide attack as low feasibility; a calibration rule or an attack vector cross-check stops that from hiding your worst case.
  • Close the loop to goals, claims and requirements. Every reduce decision becomes a cybersecurity goal; every retain or share decision becomes a justified claim with an owner and a review trigger.

What the TARA has to deliver

ISO/SAE 21434:2021, published in August 2021 and jointly developed by ISO and SAE International, describes the TARA methods in clause 15 and applies them in the concept phase (clause 9) and again whenever the item changes. The standard says what the TARA must produce; it deliberately leaves the scoring scales and matrices to the organisation. That freedom is why two suppliers can analyse the same ECU and reach different answers.

For a vehicle manufacturer the TARA also carries regulatory weight. UN Regulation No. 155 requires the cyber security management system (CSMS) to include processes to identify risks to vehicle types, including the threats in Annex 5 Part A, to assess, categorise and treat them, to verify they are managed, and to keep the risk assessment current (paragraph 7.2.2.2). Paragraph 7.3.3 asks for an exhaustive risk assessment covering individual elements, their interactions and their interactions with external systems, and paragraph 7.3.4 requires proportionate mitigations, including all relevant Annex 5 Part B and Part C mitigations. NHTSA's 2022 cybersecurity best practices point to clause 15 for the risk assessment process and say occupant and road user safety is the primary consideration when assessing risk.

In practice, an assessor looks for traceability (asset to damage scenario to threat to goal), completeness against a threat catalogue, reproducible ratings, a written reason for every rating, and defined triggers for re-running the analysis.

Step 1: Define the item and its boundary

The example item is a TCU in a passenger car. Writing the boundary down precisely decides which threats are in scope and which belong to another item's TARA.

ElementExample TCU definitionWhy it matters for the TARA
FunctionsRemote services (lock, unlock, climate), vehicle location reporting, emergency call support, OTA update download and staging for other ECUsEach function is a source of damage scenarios; OTA staging makes the TCU part of the update chain
External interfacesCellular modem to the manufacturer back end over mutually authenticated TLS; GNSS receiver; Bluetooth for the phone appThese are the remote attack surface and map to R155 Annex 5 headings 4.3.1, 4.3.2 and 4.3.5
In-vehicle interfacesCAN FD to the central gateway; automotive Ethernet to the infotainment head unitDefine the pivot paths from the TCU into safety-relevant domains
Physical interfacesDiagnostics via the gateway using UDS; a debug header on the circuit boardLocal attack paths and the R155 heading 4.3.7 physical manipulation threats
Credentials heldUnique device private key and certificate for back-end TLS; OTA package verification public key; SecOC keys for CAN messages; UDS security access secretAssets with confidentiality and integrity properties; their compromise often has fleet impact
Assumptions about the environmentThe back end authenticates devices by certificate; the gateway filters CAN traffic between domains; the vehicle owner may hold a phone keyAssumptions become cybersecurity claims or requirements on the operational environment and must be confirmed by the owners

Two boundary decisions matter most. The back end is outside the item but inside the analysis: its credentials and APIs appear as attack paths, and controls on them become requirements on the operational environment, agreed with the back-end owner. The gateway is a separate item with its own TARA; the TCU analysis records its filtering as an assumption.

Step 2: Identify assets and damage scenarios

An asset is anything whose compromise leads to a damage scenario, and each asset carries one or more cybersecurity properties: confidentiality, integrity or availability. Listing functions alone misses the credentials that make those functions trustworthy. The table shows the assets that drive the example, with damage scenarios written as consequences for road users, not as attacker actions.

AssetPropertyDamage scenarioImpact rating (S / F / O / P)
OTA package verification key and update logicIntegrityDS1: Vehicles across the fleet install unauthorised software that changes powertrain or braking behaviourSevere / Severe / Severe / Negligible
Remote unlock command pathIntegrityDS2: A vehicle is unlocked and driven away without the owner's consentNegligible / Major / Major / Negligible
Location dataConfidentialityDS3: A third party tracks the movements of identifiable ownersNegligible / Moderate / Negligible / Severe
UDS security access secretConfidentialityDS4: Diagnostic routines that alter ECU configuration are run without authorisation on many vehicles of the same lineModerate / Major / Major / Negligible
Device private keyConfidentialityDS5: An attacker impersonates the vehicle to the back end and issues or receives commands as that vehicleNegligible / Moderate / Major / Major
Emergency call functionAvailabilityDS6: Emergency call fails after a crashSevere / Moderate / Major / Negligible

The impact categories are safety (S), financial (F), operational (O) and privacy (P), each rated severe, major, moderate or negligible, with safety ratings usually aligned to ISO 26262 severity classes. Rate impact for the road user or the fleet, not the ECU: one compromised vehicle and every vehicle of a model year are different damage scenarios. Give each rating a one-line rationale, such as "Severe safety because unintended braking at speed can cause life-threatening injury".

Step 3: Derive threat scenarios and check completeness

A threat scenario is the potential cause of a damage scenario: who or what compromises which property of which asset, and how. Start per asset with STRIDE (spoofing, tampering, repudiation, information disclosure, denial of service, elevation of privilege), then check the result against R155 Annex 5 Part A, which groups threats numbered 1 to 32 under seven headings. The example yields these threat scenarios:

  • TS1 (tampering, DS1): Malicious software is distributed through the OTA chain using a stolen back-end signing or deployment credential. Annex 5 threat 12, misuse or compromise of update procedures, and the back-end headings under 4.3.1.
  • TS2 (elevation of privilege, DS1 and DS2): A remote attacker exploits a software flaw in the TCU cellular or Bluetooth stack, gains code execution and sends forged commands onto the CAN bus. Annex 5 headings 4.3.2 and 4.3.5.
  • TS3 (spoofing, DS2 and DS5): The device private key is extracted from one TCU through the debug header and used to impersonate that vehicle. Heading 4.3.7.
  • TS4 (spoofing, DS4): A shared UDS security access algorithm, recovered from one unit, is used to unlock diagnostic routines on other vehicles of the same line. Annex 5 sub-threat 11.3, malicious diagnostic messages.
  • TS5 (information disclosure, DS3): Location history is pulled from a back-end API through an over-scoped or leaked API token belonging to a partner integration. Heading 4.3.1.
  • TS6 (denial of service, DS6): Cellular jamming or a flood of malformed network messages blocks the emergency call. Heading 4.3.2.

Record excluded threats with the reason ("Not applicable: TCU has no USB port"). For deeper method background, see the Threat Modelling Guide.

Step 4: Analyse attack paths and rate feasibility

Attack path analysis turns each threat scenario into a sequence of steps an attacker must complete, either top down with attack trees or bottom up from known weaknesses. Feasibility is then rated per path, not per threat, because one threat often has a cheap path and an expensive one. ISO/SAE 21434 describes three approaches: attack potential based (built on the ISO/IEC 18045 factors), CVSS based, and attack vector based. The example uses attack potential with the factor values documented for the EVITA method, which many teams adapt.

FactorValues used in the example
Elapsed time0 (up to 1 day), 1 (up to 1 week), 4 (up to 1 month), 10 (up to 6 months), 19 (more than 6 months)
Specialist expertise0 (layman), 3 (proficient), 6 (expert), 8 (multiple experts)
Knowledge of the item0 (public), 3 (restricted), 7 (sensitive), 11 (critical)
Window of opportunity0 (unlimited, remote, no time limit) up to 10 (difficult, long physical access needed)
Equipment0 (standard), 4 (specialised), 7 (bespoke), 9 (multiple bespoke)

The organisation then maps the total to a feasibility rating. This example uses: 0 to 9 High, 10 to 13 Medium, 14 to 19 Low, 20 and above Very low. These bands are a calibration choice: fix yours in the rating guideline.

PathTimeExpertiseKnowledgeWindowEquipmentTotalFeasibility
AP1 (TS1): phish a release engineer, reuse a long-lived deployment credential, push a signed campaign4330010Medium
AP2 (TS2): find and exploit a remote flaw in the modem stack, pivot to CAN through the gateway10630423Very low
AP3 (TS3): open one TCU, read the key over an unlocked debug header1331412Medium
AP4 (TS4): reuse a published seed and key algorithm through the OBD port030104High
AP5 (TS5): use a partner API token found in a public code repository130004High
AP6 (TS6): cellular jamming near the vehicle0304411Medium

Write the reason beside every number. For AP4, for example: "Knowledge 0 because seed and key algorithms for this line have been discussed publicly; window 1 because the attacker needs brief physical access to the OBD port." Rate each path as it stands today and again after treatment; the difference is the evidence that a control is worth building.

Calibrating for remote, scalable attacks

AP2 shows the main weakness of pure attack potential scoring. A remote code execution path that takes months of expert work scores Very low, yet once developed it can be repeated against every vehicle with the same software at almost no cost. Teams handle this in one of three ways, and should document which:

  • Attack vector cross-check: rate the path with the attack vector approach as well (network-reachable paths rate higher), and take the higher of the two ratings.
  • Scalability rule: raise feasibility by one band for paths that are remote and repeatable without per-vehicle effort once developed.
  • Impact at fleet level: keep the feasibility rating but rate the damage scenario for the whole fleet, which pushes the risk value up through the impact axis.

The example applies the scalability rule, so AP2 becomes Low.

Step 5: Determine risk values and decide treatment

ISO/SAE 21434 expresses risk as a value from 1 (minimum) to 5 (maximum), set by a matrix or formula the organisation defines from impact and feasibility. The example uses this matrix, applied to the highest impact category of each damage scenario:

Impact \ FeasibilityHighMediumLowVery low
Severe5432
Major4321
Moderate3221
Negligible1111

The treatment options are to avoid, reduce, share or retain the risk. The example's acceptance rule is: values 4 and 5 must be reduced, 3 is reduced unless a justified claim is approved by the product security lead, and 1 and 2 may be retained with a recorded rationale.

PathImpact usedFeasibilityRisk valueDecisionResulting cybersecurity goal or claim
AP1Severe (S)Medium4ReduceCG1: Only software authorised by the manufacturer release process shall be installed on the vehicle
AP2Severe (S)Low (after scalability rule)3ReduceCG2: A compromise of the TCU shall not allow forged safety-relevant messages to be accepted by powertrain or chassis ECUs
AP3Major (O, P)Medium3ReduceCG3: Credentials extracted from one vehicle shall not grant access to any other vehicle or to back-end functions beyond that vehicle
AP4Major (F, O)High4ReduceCG4: Diagnostic routines that alter ECU configuration shall require per-vehicle authenticated access
AP5Severe (P)High5Reduce (operational environment)Requirement on the back end: partner tokens scoped to the minimum data, short-lived and monitored; recorded in the interface agreement
AP6Severe (S)Medium4Reduce, partly retainCG5: Emergency call shall fall back to an alternative path when the primary channel is unavailable; residual risk from jamming of all channels retained as a claim with rationale

Goals state outcomes; the cybersecurity concept chooses the mechanism. For CG4 that might be UDS authentication (service 0x29) with per-vehicle certificates, or SecurityAccess (service 0x27) with a per-vehicle secret derived in a hardware security module. NHTSA's best practices point the same way: a credential obtained from one vehicle should not grant access to other vehicles, and global symmetric keys for diagnostic access should be minimised. For the in-vehicle side of CG2, see the CAN Bus and AUTOSAR SecOC Guide.

Where identity fits in the TARA

Four of the six example paths run through a credential, which is typical of connected ECUs, so give non-human identities their own pass through the asset list.

  • Back-end machine credentials: signing and deployment credentials in the OTA pipeline, service accounts on telematics platforms and partner API tokens are attack paths into the vehicle even though they never sit in the vehicle. Give each an owner, a rotation period and monitoring, and put them in the TARA. See the Automotive Identity Security Guide.
  • Per-vehicle uniqueness: the single most effective feasibility reducer for physical paths is making every extracted secret worthless beyond one vehicle. Unique device keys, provisioned at the production line and protected in secure hardware, convert a fleet-scale damage scenario into a single-vehicle one.
  • Certificate lifecycle: device certificates expire and get revoked; include the back-end certificate authority and its revocation process as an asset with integrity and availability properties. See the Machine Identity, PKI and Certificate Lifecycle Guide.

Running the TARA as a living work product

R155 asks for the risk assessment to be kept current, and NHTSA recommends periodic re-evaluation as threats change. Treat the TARA like code:

  1. Baseline: complete the concept phase TARA, agree rating guidelines and the risk acceptance rule, and put both under configuration control.
  2. Link to design: trace each goal to requirements and each requirement to verification evidence, so a design change shows which risks it touches.
  3. Define triggers: re-run affected parts on a design change, a new vulnerability in a component (NHTSA recommends keeping a database of hardware and software components per ECU for exactly this), new published attack research, or a field incident.
  4. Re-rate after verification: once controls are tested, re-score the affected paths and record the residual risk.
  5. Share with suppliers: pass the relevant goals and assumptions to suppliers through the cybersecurity interface agreement, and receive their component-level analyses back.
  6. Review residual claims: every retained or shared risk gets a review date and a named owner.

Assurance rigour is a separate question from risk. ISO/SAE 21434 describes cybersecurity assurance levels (CAL) in an informative annex, and ISO and SAE have been developing ISO/SAE PAS 8475 on CAL and targeted attack feasibility (TAF); the ISO project page lists it at stage 60.00 (under publication) from September 2026, so check whether it has been published before relying on it.

Measuring TARA quality

  • Traceability coverage: percentage of risks with value 3 or above that trace to a goal or claim and onward to verification evidence (target 100%).
  • Justification coverage: percentage of feasibility factors and impact ratings with a written rationale.
  • Currency: days between a trigger event (design change, new vulnerability) and the updated TARA.
  • Escape rate: attack paths found in penetration testing or the field that the TARA did not contain.
  • Rating consistency: agreement between two engineers independently rating a sample of paths with the same guideline.

Framework mapping

TARA activityISO/SAE 21434 and UN R155NIST CSF 2.0 and SP 800-53 Rev. 5
Asset, damage and threat scenario identificationClause 15; R155 paragraph 7.3.3 and Annex 5 Part AID.RA-04; RA-3 Risk Assessment
Attack path analysis and feasibility ratingClause 15ID.RA-05; SA-17 Developer Security and Privacy Architecture and Design
Risk treatment and goalsClause 15 and clause 9; R155 paragraph 7.3.4, Annex 5 Parts B and CID.RA-06
Supplier analyses and interface agreementsClause 7GV.SC-05; SR-3 Supply Chain Controls and Processes
Re-assessment after vulnerabilities and testingClause 8; R155 paragraph 7.2.2.2ID.RA-01; RA-5 Vulnerability Monitoring and Scanning; SA-11 Developer Testing and Evaluation

Common failure patterns

  • Function-only asset lists: listing "remote unlock" but not the keys and tokens that authorise it, so the highest-impact credential paths never appear.
  • Unexplained scores: feasibility totals with no reason per factor, which an assessor cannot check and the next engineer cannot update.
  • Single-vehicle impact thinking: rating a shared secret or a remote flaw as if it affected one car, when the same attack works on the whole fleet.
  • Back end left out: treating the cloud as someone else's problem, so OTA pipeline credentials and partner tokens are missing from a vehicle TARA that depends on them.
  • Goals written as mechanisms: "use AES-128 CMAC" as a goal locks in a design before the concept is agreed and hides the outcome the reviewer needs to verify.
  • A static document: no triggers, no re-rating after testing, and claims with no owner or review date.

Practitioner checklist

  • Write the item definition with functions, interfaces, credentials held and assumptions about the environment.
  • List credentials (device keys, OTA keys, SecOC keys, diagnostic secrets, back-end tokens) as assets with properties and damage scenarios.
  • Rate impact at fleet level where the attack scales, with a one-line rationale per category.
  • Check threat scenarios against STRIDE per asset and R155 Annex 5 Part A, and record exclusions with reasons.
  • Rate feasibility per attack path with a written reason for each factor.
  • Document a calibration rule for remote, repeatable attacks.
  • Publish the risk matrix, the acceptance rule and the rating guideline under configuration control.
  • Turn every reduce decision into an outcome-based cybersecurity goal and every retain or share decision into a claim with an owner and review date.
  • Define re-assessment triggers and re-rate paths after verification.

Standards and references

Related NHI Mgmt Group resources: ISO/SAE 21434 Automotive Cybersecurity Engineering Guide · Automotive Cybersecurity Guide · Automotive Cybersecurity Regulations Compared · Connected Fleet and Telematics Security Guide · Debug Interfaces Security Guide

PDF for subscribers

Take this guide with you

Every section, table and checklist in one PDF, with links to the sources and related guides. Subscribe to our weekly NHI & AI Identity Journal and we’ll email you a download link.

Free weekly newsletter

Subscribe to the NHI & AI Identity Journal

The latest on NHI and Agentic AI security – articles, research, breaches, news and events every week.

Bonus 33% off our NHI Course when you subscribe.

NHIMG Editorial Note
Researched and written by the NHI Mgmt Group editorial team using AI tools, under the editorial direction of Lalit Choda. Facts are drawn from the sources listed on this page and checked by our automated quality process. Last updated 14 November 2026. Spotted an error? Use our contact page.
NHI Mgmt Group, the independent authority on Non-Human Identity, IAM and Agentic AI security. nhimg.org

PDF Downloads for Newsletter Subscribers

Get the guide

Enter your details and we’ll email you a download link.

The full guide is free to read online. The PDF is a thank you for subscribing.

Fill in all fields marked * and tick the PDF box to continue.