FMEAYou Bought an ASIL-B Storage Device. Is Your Safety Analysis Done?-compute SSD example through HARA, FTA, FMEA and TARA
01Summary
02What the Article Tells Us and What It Doesn't
In a November 5 article, the German electronics publication all-electronics argued that zonal architecture, OTA and V2X in SDVs are turning storage into a part where safety and security overlap. Its example, a Silicon Motion automotive SSD controller, is described as certified to AEC-Q100, ISO 26262 ASIL-B and ASPICE CL3.
Keep three things in mind. The article states that it is based on material from that company. The wording we found in the 2023 product announcement goes as far as "ASPICE compliance", and we have not seen the original certificates. The article gives no failure rates or specific safety mechanisms, and it describes ISO 21434 as conformity, not certification.
One term in the article is still useful: SEooC (Safety Element out of Context), a safety element developed without a specific vehicle in mind.
03What a Component Certificate Hands Over and What It Leaves
A supplier builds an SEooC part around assumed conditions of use. Those assumptions are written in the safety manual and passed to the integrator. If they do not hold in your vehicle, you cannot reuse the part's safety argument as it stands.
| What the certificate shows | The question left for the vehicle side |
|---|---|
| The supplier's development process and safety mechanisms were assessed against a target level | What is the ASIL of the safety goals that depend on this part? |
| Operation is supported under defined temperature, power and usage conditions | Do our mounting location, power supply and write volume stay within them? |
| The part can diagnose its internal failures | Does the vehicle receive that result and respond safely? |
| Stored data is encrypted | Who manages the keys, and who stops tampering through legitimate paths? |
Let's follow these questions through a fictional example.
04Example: A Central-Compute SSD Sharing One Controller
This configuration was made for education. One SSD in the central computer is shared by driving assist, infotainment and OTA virtual machines. We assume the controller is certified to ASIL-B. Driving assist reads map data from the SSD, and host software checks integrity with a checksum.
All ratings, scores and failure rates below are assumptions made for illustration. They are not values from a real product.
HARA: The Part's Level Is Not the Safety Goal's Level
| Malfunction | Hazardous event | S | E | C | ASIL |
|---|---|---|---|---|---|
| Provides outdated or wrong map data as if it were valid | Driving assist on a highway leaves the lane because of an old map | S3 | E4 | C2 | C |
| Does not provide map data when needed | Driving assist on a highway switches off without warning | S2 | E4 | C2 | B |
The first safety goal is ASIL C, and the controller is ASIL-B. So the design team chooses ASIL decomposition: ASIL B(C) for the SSD side and ASIL A(C) for the host integrity check. In the table, the levels add up. But ASIL decomposition comes with a condition: the decomposed elements must be sufficiently independent.
FTA: Two "Independent" Elements Were Tied to One Controller
Suppose the checksum used for the integrity check is stored on the same SSD as the map data. If a NAND error corrupts only the data, the checksum catches it. But if the controller returns a previous version of a block, the data and the checksum come back old together. They match each other, so the check passes. A healthy check program makes no difference.
The T0 structure has three cut sets.
| Cut set | Order | Meaning |
|---|---|---|
| Controller firmware returns a previous block version | First | This one failure causes the top event |
| Controller internal mapping table error | First | This one failure causes the top event |
| NAND data error + host integrity check failure | Second | The dual structure the designer intended |
The ASIL decomposition table shows only the second-order case. The two first-order cases appear only once you draw the FTA.
The Fix: Move the Reference Outside the Controller
Keep a signed reference value and a version number in a separate secure store. Now, if the controller returns an old block, it no longer matches the value in the separate store and gets rejected.
| Structure | First-order cut sets | Second-order cut sets |
|---|---|---|
| T0 Reference on the same SSD | 2 | 1 |
| T1 Reference in a separate store | 0 | 3 |
The part stays the same. Only the vehicle-side structure changed. Because the failure rates are assumptions, this article compares cut set structure, not calculated values.
FMEA: What Breaks the Supplier's Assumptions Sits on the Vehicle Side
| Failure cause | Failure mode | Failure effect | S | O | D | AP |
|---|---|---|---|---|---|---|
| Mounting temperature exceeds the supplier's assumed range | Data retention drops and uncorrectable errors increase | Driving assist uses wrong map data | 9 | 4 | 6 | H |
| Hold-up time is too short when power is lost during an OTA write | Metadata corruption makes the stored image unreadable | Some vehicle functions do not start after an update | 7 | 5 | 7 | H |
| Log volume exceeds the supplier's assumption | Write endurance runs out and the drive turns read-only | Some vehicle functions do not start after an update | 7 | 6 | 6 | H |
A component certificate cannot answer any of these three causes. Checking the part's temperature grade tells you nothing until you measure the actual mounting location, and as OTA adds features after SOP, write volume drifts from the original calculation. That is why test plans need to move from the bare part to in-vehicle conditions.
TARA: What AES-256 Stops and What It Doesn't
| Threat | Attack potential total | Feasibility | Impact | Risk level | Decision |
|---|---|---|---|---|---|
| Remove the SSD and read NAND directly | 42 | Very Low | Major | 1 | Retaining |
| A compromised infotainment VM tampers with map data over the shared path | 18 | Medium | Severe | 4 | Reducing |
| Extract a shared key from one vehicle and decrypt data across the model | 22 | Low | Major | 2 | Reducing |
Encryption of stored data directly stops the first threat. The second requests a write with legitimate rights, so it gets through regardless of encryption. The third is about key management, not cipher strength. So the countermeasure for the highest-risk threat becomes "separate storage areas per virtual machine and verify signatures before use". That signature check leads to the same design as the reference separation found in the FTA. Here, safety analysis and security analysis point to the same countermeasure.
05Five Things to Check When You Receive a Component Certificate
| What to check | Where to look | What the example showed |
|---|---|---|
| The product, firmware version and scope on the original certificate, and whether it is a certification or a conformity claim | Certificate, supplier documents | Sources mix "certified" and "compliant" |
| The ASIL of safety goals that depend on the part | HARA | Part ASIL-B, safety goal ASIL C |
| Resources shared by decomposed elements | FTA common causes | Two first-order cut sets from one controller |
| Gaps between the safety manual's assumptions and your design | FMEA failure causes | Temperature, hold-up time, write volume |
| Scope of encryption, key management, access control on shared paths | TARA | Encryption stops one threat out of three |
06How to Do This in VisualPro, and the Limits
VisualPro keeps HARA, FTA, DFMEA and TARA in one project. We entered this example as four modules in one project, and the tool calculated ASIL, AP, risk levels and cut sets.
VisualPro does not replace everything. Real failure rates have to come from supplier data and testing. Judging whether a certificate is valid, or whether the independence behind an ASIL decomposition is sufficient, belongs to engineers and assessors. This example does not cover part-level FMEDA.
07Frequently Asked Questions
Q1Can an ASIL-B certified part be used in an ASIL C or D function?
In some cases, yes. ASIL decomposition is the typical route. The decomposed elements must be sufficiently independent, though. Shared resources can create a single point of failure in the FTA, as in this example, so check for them first.
Q2What do we still have to do with an SEooC certified part?
Compare the safety manual's assumptions with your vehicle's conditions one by one. If an assumption does not hold, change the design or work it out with the supplier. Connecting the part's diagnostic results to a safe vehicle response is also the integrator's job.
Q3Does AES-256 encryption on the storage device lower risk in TARA?
It helps against threats that remove the storage medium and read it. Tampering through legitimate paths and shared key leaks must be assessed separately. Splitting attack paths by threat shows how far encryption actually reaches.
References
· all-electronics, "Speicher wird zum Sicherheitsfaktor" (2026-11-05) — Read article
· Silicon Motion, August 2023 product announcement — View source
· Intertek, "Exploring Safety Elements out of Context" (2026-01-15) — View source
· Renesas, applying ISO 26262 to SEooC software — View source
· ISO 26262 Road vehicles — Functional safety — ISO page

