Free tool
What can a stranger with a handheld see?
Five tags on a rack, each protected a different way. Pick what the stranger is trying to do and see what they get, what your back end can still catch, and what the reader itself needs.
Stand in the aisle
Tag 1
The stranger gets
Your back end sees
Tag 2
The stranger gets
Your back end sees
Tag 3
The stranger gets
Your back end sees
Tag 4
The stranger gets
Your back end sees
Tag 5
The stranger gets
Your back end sees
The reader is the bigger surface
Tags answer anyone. The reader answers your network.
How it works
What you’re looking at
A tag has no idea who is asking. A handheld in the aisle gets the same answer your dock door does. So the question is never “can they read it” but “what did they get, and what can you do about it”.
The five tags are five answers to that. Nothing set is what the standard calls unprogrammed: both passwords at zero, nothing locked. An access password locks the EPC bank and the passwords against writes, and the EPC still reads. Impinj’s Protected Mode keeps the tag silent until a reader sends the password. NXP’s UCODE DNA reads normally and can also answer an AES-128 challenge, so a copied number has no key behind it. A killed tag is gone, for them and for you.
Each card says what the stranger walks away with, and what a back end that records the EPC/TID pair at encoding and keeps a last-seen location per tag would notice. The TID is locked at the factory, so a blank tag keeps its own. Recording the pair is cheap, and it catches the ordinary copy.
It doesn’t catch every copy. A 2017 IEEE RFID keynote lists the TID as basic protection with no defence against emulators or against chips with a writeable TID. If the thing you tag is worth that effort, tags that authenticate are the next step, and then the check has to actually run.
The reader is the bigger surface
A tag can tell a stranger a number. A reader can tell them everything it sees. Many are small computers on your network: Impinj’s R700 runs Linux, and a Nedap integration guide lists its default login as root and impinj. Most speak LLRP, which the IANA registry puts on TCP port 5084. The checklist is eight lines. It lives in the link, so tick what’s done and send it on.
What it leaves out
Methods. The page says what each protection stops and what it doesn’t, never how anything is done. It’s a map of the Gen2 standard and two vendors’ documents, not a security assessment, and nothing here is a guarantee that a deployment is safe. A 32-bit password is a small space. Get a real review before you rely on any cell in the matrix.
It also leaves out item-level privacy, which is its own issue: a tag on a jacket answers in the street as readily as in the stockroom. The Gen2v2 Untraceable command lets a tag hide part of its memory or cut its range, and Impinj’s Short-range Mode cuts read range by more than 90%. Killing at the till is the older answer.
Where this comes from
- GS1 EPC Gen2 v2.0.1: 32-bit access and kill passwords, both zero until set; no password Kill while the kill password is zero; a zero access password goes straight to the secured state; a locked password can’t be read or written without the access password, unless permalocked; the EPC and TID banks, once locked, still read; the spec recommends unique kill passwords and against a zero access password.
- Impinj, Protected Mode page: the tag doesn’t respond unless the reader sends the correct password. Listed for the M700 and M800 series.
- NXP UCODE DNA brochure and product page: AES authentication to ISO/IEC 29167-10, two 128-bit keys in the chip’s guarded memory, a 96-bit TID with a 48-bit serial.
- Martinez, IEEE RFID 2017 keynote: what a TID does and doesn’t defend against.
- IANA: port 5084 is the Low-Level Reader Protocol.
- Impinj’s blog (the R700 runs a Linux kernel) and Nedap’s R700 guide (the default login, and the prompt to change it from firmware 9.0).
The matrix is in src/config/rfid-security.ts on the site’s repo, with each claim’s source beside it.