Attack Encrypted NRF52833-QDAA-R Flash: QFN40 512 KB Recovery
The NRF52833-QDAA-R is the smallest package in the nRF52833 family, a QFN40 body of 5 x 5 mm with 18 GPIO that still carries the full Bluetooth 5.1 radio, the USB 2.0 full-speed controller and the direction-finding hardware. When the image is sealed and the board is the only copy, the work is to attack encrypted NRF52833-QDAA-R flash and bring the contents out intact. The catch on this part is that the die is packed close under the compound, so the physical route has to be controlled rather than brute forced.
The densest logic in the smallest footprint
Fitting a complete radio SoC into 5 x 5 mm is the reason this variant exists, and it is also where the difficulty concentrates. The part carries 512 KB of flash and 128 KB of RAM behind a Cortex-M4 with FPU at 64 MHz, scoring 217 EEMBC CoreMark from flash at 52 uA/MHz, or 38 uA/MHz from RAM. The radio reaches +8 dBm with transmit power adjustable from -20 dBm in 4 dB steps, holds sensitivity of -96 dBm at 1 Mbps and -103 dBm in 125 kbps long-range mode, and reports RSSI at 1 dB resolution. Recovering an image from a die that small leaves less physical margin to work with at every stage.
Where the protection actually lives
The nRF52 does not use the STM32-style three-level readout protection, so tools built for that model find nothing familiar here. The lock sits in the User Information Configuration Registers as APPROTECT, alongside a control access port, CTRL-AP, that stays available. Writing Enabled, 0x00, to UICR.APPROTECT closes the debug access port in hardware, and the hardware fetches that value at reset, so the state survives a power cycle with no firmware cooperation. Around it sit a flash access control list, secure boot readiness and the debug control and configuration blocks. Reading the UICR is therefore part of understanding the part, not a formality at the end of the job.
The two-step lock and the revision window
Nordic splits the lock into a hardware step and a firmware step, and the gap between them is where the openings are. Firmware is expected to write APPROTECT.FORCEPROTECT as Force, 0x00, at startup, and if that firmware step is skipped the device can look locked without being fully protected, which is often the first thing a careful read of the configuration area reveals. Silicon revisions with the improved two-step APPROTECT, among them nRF52833 revision 2 under build codes Bx0, tightened that check, and on those parts a precisely timed voltage glitch on the supply during boot can slip past it without triggering the erase. The unlock path runs the other way: a CTRL-AP ERASEALL wipes flash, UICR and RAM, code compiled without ENABLE_APPROTECT writes HwDisabled, 0x5A, and after a reset the firmware writes APPROTECT.DISABLE to open the port.
Both APPROTECT.DISABLE and FORCEPROTECT are reset by a pin reset, a power or brownout reset, a watchdog reset, or a wake from System OFF, so any such window has to be caught in a single pass rather than held open.
Ranking the attack surfaces
Against a locked NRF52833-QDAA-R the practical surfaces rank by effort and by what they preserve. The cheapest route that still keeps the image is the one worth trying first.
| Surface | Preserves image | Effort | When used |
|---|---|---|---|
| Supply glitch during boot | yes | medium | revision 2, build codes Bx0 |
| Targeted fault at reset | yes | medium | partly protected units |
| Etch-back and direct die read | yes | high | fully locked, no register route |
| CTRL-AP ERASEALL and reload | no | high | only when loss is acceptable |
Etching back to the silicon
When the register-level routes are closed, the answer is the physical one. The QFN40 body leaves little compound margin, so the mould has to come away in controlled passes while the board keeps running, and the etch has to stop on the die surface without cutting into the pad frame or the exposed thermal pad. Once the silicon is exposed and still live, the array can be addressed directly on the die rather than through the debug port, which sidesteps the access port lock entirely and is the surest way to attack encrypted NRF52833-QDAA-R flash when the register doors are welded shut. Our notes on package decapsulation and on supply glitching describe the technique in detail.
What you get back
You receive the flash binary image of the NRF52833-QDAA-R with the UICR region and the protection state documented, a disassembly-based source reconstruction built from the vector table and the peripheral register map, and a verification report recording how the image was obtained and how it was checked. Where the aim is a working replacement rather than an analysis, the same verified read supports a protected flash duplication and the matching QDAA-R7 source rebuild. The same bench also handles the QIAA-R hex image copy and the locked STM32G474 program.