What a permissioned blockchain can and cannot prove: a CTO's checklist
I'm Pratyush Shivam, CTO of Konfidio in Berlin. I lead the technology behind TrustedTrucks, a freight platform where the shipper, the forwarder, the carrier and the consignee work from one shared shipment record, and whe
I'm Pratyush Shivam, CTO of Konfidio in Berlin. I lead the technology behind TrustedTrucks, a freight platform where the shipper, the forwarder, the carrier and the consignee work from one shared shipment record, and where shipment documents are fingerprinted on a permissioned blockchain.
The question I get most often, from engineers and from customers, is some version of "so the blockchain proves it?" The honest answer is: it proves some things very well, and other things not at all. Being precise about which is which is most of the job. This is the checklist I use.
What it can prove
1. That a document has not changed since it was anchored. A fingerprint (a hash) of the document is written to the chain. If anyone later presents a file that hashes differently, it is not the same file. This is the strongest claim the system makes, and it is a narrow one.
2. When it was anchored. The entry sits in a block with a position in an ordered history. You can say "this version existed at this point", and nobody can quietly backdate it.
3. Which participant anchored or approved it. On a permissioned network the participants are known companies with known identities. An approval is tied to the account that made it, so "who approved this version?" has a checkable answer months later.
4. That every party is looking at the same history. The forwarder and the carrier don't each keep their own copy that drifts. They read the same record, so a dispute starts from shared facts instead of competing spreadsheets.
What it cannot prove
1. That the document was true. If a delivery note was wrong when it was signed, the chain faithfully preserves a wrong delivery note. A fingerprint is not a fact-check.
2. That the goods match the paperwork. The chain sees data, not pallets. A proof of delivery says a person accepted a delivery. It doesn't count the boxes.
3. That the right person pressed the button. It proves which account approved something. Whether that account was used by the person it belongs to is a question for sign-in, roles and device security, not for the ledger.
4. Anything about data that never reached it. If a step happens over email and never touches the shared record, the record is silent about it. A shared record is only as complete as the process that feeds it.
What this means for design
Once you accept those limits, the design decisions get clearer:
- Put people at the points of truth. Approvals happen where a human can actually check something: at the dock, on the phone, at the moment of delivery.
- Keep the chain underneath. A dispatcher should see "approved", never a wallet or a hash. The guarantees are for when something is disputed, not for everyday screens.
- Let evidence flow forward. An accepted proof of delivery feeds the invoice directly, so the invoice is backed by the same evidence everyone already agreed on.
- Say the claim exactly. "This is the version that was approved, when, and by whom" is a claim you can defend in any meeting. "The blockchain proves the delivery happened" is not.
The checklist, short
Before saying "the chain proves it", ask:
- Are we claiming integrity, timing or attribution? Those it can prove.
- Are we claiming truth, physical reality or a person's identity? Those need something else.
- Did the step actually touch the shared record?
- Would the claim survive a sceptical customer reading it word for word?
Photo: at the podium at the Software Technology Parks of India in Ranchi, the city I'm from.
More of my writing, films and projects are at pratyushshivam.com.
Originally published by Dev.to Security. Aggregated on AIWithGhost for educational purposes β full credit and traffic to the original publisher.