Flipper One on a company network: EUR 11 million in pledges, 339 access ports and no port authentication
On 17 November 2026 the Kickstarter campaign for the Flipper One started, a small Linux machine with two gigabit Ethernet ports. On 18 November the page showed 11,386,689 euros in pledges from 23,421 people, measured close to 22:15 UTC. Delivery is stated by the vendor as July 2027. This is not a product review, I never had the device in my hand. What I have is a measurement of our own access layer, and that is more uncomfortable than any number on the product page.
802.1X port authentication means a switch port releases traffic only after the device has identified itself to an authentication server, in practice over RADIUS. The standard is IEEE 802.1X-2020, and RFC 3580 describes how RADIUS is used with it. Without proof the port stays closed, lands in a guest VLAN, or gets a helper profile. Judged is the device, registered as a certificate or as a MAC address.
What was funded, each figure in its own currency
These numbers move in three currencies. The rows are transcripts from the vendor pages and the press, as of 18 November 2026 (the unit sits in the Stated as column). A newer state is not available: the Kickstarter page answers automated requests with 403.
| Quantity | Stated as | Where it stands |
|---|---|---|
| Flipper One pledges | 11,386,689 EUR, 23,421 backers, 28 days left | Kickstarter project page, display currency per account |
| Pledges on the morning of 18 Nov | "almost ten million euros", over 20,000 backers | heise, 18 Nov 2026 |
| Base unit | 399 USD, listed as 25 % below an expected 550 USD list price | Rewards list, prices in USD |
| Base unit in euros | 364 EUR, full package with all modules about 691 EUR | heise, quoted as a conversion |
| All three modules together | 759 USD, on their own 59 / 149 / 219 USD | Rewards list |
| First million | 10 minutes after launch | heise, no source given in the text; not checked against the Kickstarter page |
| Flipper Zero campaign 2020 | 4,882,784 USD, 37,987 backers | 2020 campaign; rounded 4.88 million USD at Wikipedia |
| Same Zero campaign | "about 4.3 million euros" | heise, quoted as a conversion |
Kickstarter shows pledges in the account's display currency, the prices are in US dollars, heise converts into euros. I add nothing across currency borders.
The record claim needs the same checking. The Guinness entry Fastest $1 million pledged for a crowdfunded project still lists Kingdom Death: Monster 1.5 at 19:00 minutes from 25 November 2016 on 19 November 2026, with Pebble Time before it at 49 minutes. The title names US dollars, and counted are pledges, not money received. heise gives the Flipper One's ten minutes without a currency, and Guinness itself notes that records do not appear online immediately. Whether Flipper Devices files the ten minutes is open (heise).
What the hardware does, and where the page stays quiet
The tech specs and the product page give no RAM capacity; 8 GB LPDDR5 appears only in the press (It's FOSS).
The vendor separates the devices: the Zero is the handheld for radio and cards (NFC, RFID, sub-GHz, infrared), the One covers the network and computer layer. The tech specs name a Rockchip RK3576 with 4 Cortex-A72 and 4 Cortex-A53 up to 2.2 GHz, a Mali G52 MC3, an NPU at 6 TOPS @INT8, 64 GB of UFS 2.2, Wi-Fi 6E with 2×2 MIMO (MediaTek MT7921AUN), two gigabit ports (PHY Realtek RTL8211F-CG) and HDMI 2.1 up to 4K at 120 Hz. The NPU figure is the vendor's word, in tera operations per second on 8-bit integers, not measured by us. The same RK3576 appears on the product page of the Violoop, which states a different NPU figure.
Three points where the spec gives no number:
The memory is open. The spec page says TBD for RAM; capacity and type (LPDDR4 or LPDDR5) are decided after validation and on market conditions. The campaign's risk section names the reason: "we cannot rule out adjusting some product specifications, including reducing the amount of RAM." That quote could not be re-checked against the Kickstarter page (403). The press does name a figure: It's FOSS writes 8 GB LPDDR5, and BleepingComputer quotes the vendor saying RAM and UFS make up about half of the hardware cost. The spec page and the product page carry no RAM figure; anyone comparing the 8 GB compares a press number.
The SDR range is missing from the vendor documents. heise gives 24 to 1,700 MHz for the SDR module. On the tech-spec page and the product page I find 0 hits, and BleepingComputer writes that the full module specifications are still to be published. A figure from the press does not become a vendor figure.
Power over the network cable is not stated. For Power over Ethernet I find 0 hits on the tech-spec page and the product page. The device charges over USB-C1 with USB Power Delivery. The slot on the back is an M.2 socket (key B) with PCIe 2.1 x1, USB 3.1 and SATA3; the 500 MB/s raw behind that is derived.
Open source is not the same as independently checked
The stated goal is a machine with a mainline-capable kernel and no proprietary blobs, and that is the most valuable sentence on the product page because it sets lifespan and verifiability. The vendor names Collabora as its partner for the kernel work, and Collabora confirms it in a post of 21 May 2026. The licence situation in the public repos is measured: the MCU firmware repo is mostly MIT with a GPL-3.0-or-later share and CC-BY-SA-4.0 for graphics and documentation. The repos for build scripts, docs and testing carry no licence file. The same organisation also runs rkbin, a repo of Rockchip binaries whose licence text allows use and redistribution but forbids decompiling and reverse engineering. Whether the shipped image needs those files, none of the sources says. Anyone who lets such a device into a company network needs a capture point for firmware state and licence paths, and nobody offers that today.
Our access layer lets every one of these devices through
Now the own measurement. Our switches write their running configuration to a server by TFTP every night; evaluated is the copy from 11 November 2026, the configs themselves are older. I counted the interface stanzas with a fixed access VLAN on eleven devices, then searched the same files for every instruction on port authentication, 802.1X, dot1x, RADIUS and host mode. Result: 852 interface stanzas, of them 339 access ports with a fixed VLAN, of them 0 with port authentication. Unit: switch interface. Carrier: the nightly config copies of our 11 access and core switches, measured on 18 November 2026. The spread runs from 122 access ports on one Cisco to 7 to 28 on the others. The only hits for the word "authentication" across eleven files are two SNMP lines about locked accounts.
The finding is the rule behind the port: on 339 ports the switch decides nothing about the device, it carries a VLAN number a human typed in years ago. A device someone carries in inside a briefcase gets a VLAN from a free wall socket, and with it a route to everything those VLANs reach. The 339 ports therefore concern any unknown Linux box, the radio side is a side issue. Two posts from our archive show the same gap from another angle: A coffee machine sent 1 TB in 10 days: put IoT on its own VLAN, on IoT devices, per-client traffic counters and everyday segmentation, and An AI agent as a USB keyboard: Violoop checked, different mechanics, the same blindness in the inventory. The groundwork on when VLANs become mandatory for us and how we build them is in Network Basics: VLANs & Jumbo Frames.
Limits of the measurement: compared are configuration texts, the runtime state of those ports is unverified, and so is the wireless side with its own guest logic. The statement covers the access layer of the eleven devices and the evaluated state of the TFTP copies from 11 November 2026. The traffic watchdog on the measuring host listens on its own interface and does not see the access layer.
What would have to change, and what that costs
The requirement is older than the device: NIST SP 800-53 Rev. 5 asks in control IA-3 to uniquely identify and authenticate devices before a connection is established. The control's guidance names IEEE 802.1X with EAP and a RADIUS server with EAP-TLS as means, and it lists MAC and IP addresses as known, shared information. The strength of the check follows the protection need of the system.
Port authentication on 339 ports is not a switch you flip. It needs a RADIUS service, device or user certificates, a rule for everything unauthenticated (closed, guest VLAN, printer exception by MAC) and an exception for devices that do not speak 802.1X. The fallback to the MAC address is weak: the LANCOM reference manual says it is easy to fake and does not protect against malicious attacks. A gap remains even then, and two Ethernet ports bear on it: a device between an already logged-in computer and the port can ride on that login. The nac_bypass project has collected tools for this for years. Whether the Flipper One gets used for it is not documented; the two ports are the technical precondition. On top, part of our fleet has no guest agent reporting inventory and patch level, so the stock is missing there. The first sensible step is recognition at the port: a state compared regularly against the device inventory, where a new MAC on an access port triggers a report. With identity per port, "block the port" turns into "note the port" on its own. I am not naming an effort figure, we have not done that work and any number would be invented. Documented is the surface: 339 ports on eleven devices.
The measurement matters for any business with fixed access VLANs; the figures above come from a single network. For us this is WATCH, with no pilot. The device is not purchasable, the spec stays open on memory and weight as well as on the radio module, and delivery is stated from July 2027. Reassess when three things come together: a shippable device, a final spec, and a free access layer on which we would actually notice an unknown Linux box. Anyone who does not want to leave that to chance in their own network finds the rest of the series under Security and data protection. For measuring in your own network without your own RADIUS project, our IT services are the frame.
Further Reading
- Kickstarter project page Flipper One: pledges, backer count, every reward tier and the 2027-07-01 delivery date per tier; read on 18 November 2026 around 22:15 UTC. The page answers automated access with 403; on content it is the primary source for the table rows on pledges and prices.
- Flipper One product page and tech specs in the docs portal: hardware values, the fields left open, and the mainline goal; specs last changed 17 November 2026.
- Collabora: Opening up the RK3576: the partnership for the open-source software side, 21 May 2026.
- The flipperdevices organisation on GitHub: licence files per repo, as of 19 November 2026, including the rkbin repo.
- heise article accompanying the video: the record claim, the euro figures and the SDR range, 18 November 2026.
- The video by heise and c't: Flipper One, Millionen-Hype um den neuen Hacker-Computer: the trigger for this post.
- BleepingComputer on the campaign launch and It's FOSS: campaign dates, price, delivery estimate July 2027.
- Guinness entry for the fastest million in a crowdfunded campaign: leading entry and counting method, as of 19 November 2026.
- IEEE 802.1X-2020 and RFC 3580: IEEE 802.1X RADIUS Usage Guidelines: the standard and RADIUS usage including VLAN assignment by tunnel attribute.
- NIST SP 800-53 Rev. 5: control IA-3, device identification and authentication.
- LANCOM reference manual: 802.1X authenticator on Ethernet ports: MAC-address fallback, single-host and multiple-hosts modes.
- scipag/nac_bypass: public script collection for bypassing 802.1X.
What does 802.1X mean on a switch port?+
IEEE 802.1X-2020 describes port-based network access control: a port releases traffic only after the device has identified itself to an authentication server. The switch asks over RADIUS, and the server can hand back a VLAN in its answer (RFC 3580). Without valid proof the port stays closed or lands in a guest VLAN.
How do you keep unknown devices off free network sockets?+
With 802.1X port authentication: the switch releases the port only after the device has identified itself to the RADIUS server. That takes a RADIUS service, certificates or accounts, and a rule for devices without 802.1X such as printers. Falling back to the MAC address for them is, per the LANCOM reference manual, easy to fake and does not protect against malicious attacks. None of our 339 access ports has such a check.
Does 802.1X help against a device like the Flipper One?+
Against an unknown device without a certificate it helps: the port stays closed or the device lands in the guest VLAN. Against a device sitting between an already authenticated computer and the port, 802.1X alone is not enough. Public tools such as nac_bypass have covered that for years, and the Flipper One has two Ethernet ports. That it gets used this way is not documented; it is the technical possibility.
When can the Flipper One be delivered?+
The Kickstarter campaign runs from 17 November to 17 December 2026, with delivery stated from July 2027. heise reminds readers that the Flipper Zero reached its backers more than a year late. No date guaranteed by the maker and its supply chain exists.
senn-tech