What Is ARM Cortex-M Firmware Readout?
ARM Cortex-M firmware readout is the process of extracting the binary contents of an MCU’s internal flash memory—typically through debug interfaces (SWD/JTAG) or alternative techniques—when standard read access has been restricted by the chip’s protection features. Engineers pursue firmware readout for authorized purposes: recovering lost source code from legacy products, restoring production lines after a departing developer, duplicating your own proprietary boards, or maintaining obsolete equipment where the original firmware is the only path to continued operation.
The Cortex-M family spans hundreds of silicon implementations from STMicroelectronics, NXP, Microchip, Infineon, Nordic, Nuvoton, Renesas, Maxim, and others. Each vendor layers its own flash-protection scheme on top of ARM’s CoreSight debug architecture. That vendor-specific layer is what makes ARM Cortex-M firmware readout a per-family engineering challenge rather than a universal procedure.
How Does ARM Cortex-M Debug Protection Work?

ARM’s CoreSight specification defines a Debug Access Port (DAP) that connects an external debugger to the processor’s internal buses. When protection is disabled, a debugger can freely read every address in the flash and SRAM map. Vendors add protection by gating or disabling this access at one or more levels.
Common Protection Tiers
| Protection Level | Typical Behavior | Readout Feasibility |
|---|---|---|
| Level 0 (Unprotected) | Full debug access; flash readable via SWD/JTAG | Trivial—standard tools |
| Level 1 (Read-Out Protection) | Debug port active but flash reads blocked; mass-erase required to unlock | Moderate—vendor-specific bypass may exist |
| Level 2 / Permanent Lock | Debug port permanently disabled; fuses blown irreversibly | Difficult to impossible without invasive methods |
| TrustZone / Secure Boot | Hardware isolation of secure world; cryptographic boot chain | Very difficult; depends on implementation flaws |
Level 1 is the most commonly encountered scenario in production boards. It blocks casual readout but does not destroy the debug infrastructure. This is where the majority of successful ARM Cortex-M firmware readout work happens.
Family-Specific Protection and Readout Realities
No two vendors implement protection identically. Below is an honest, engineer-level summary of what our lab sees across the most common Cortex-M families.
STM32 (STMicroelectronics)
STM32 parts use a three-tier scheme: RDP Level 0, 1, and 2. Level 1 disables flash read via the debug port but keeps the debug connection alive for limited operations. Level 2 permanently disables SWD/JTAG by blowing option bytes—once set, even ST’s own tools cannot reconnect. For RDP Level 1, several known voltage-glitching and fault-injection techniques have been publicly documented. Success rates depend on the specific die revision; newer STM32H7 and STM32WB55 parts with dual-core secure enclaves are significantly harder than older STM32F1 or STM32F4 devices.
NXP LPC & Kinetis
NXP’s LPC series uses Code Read Protection (CRP) with levels 1–3 plus a “NO_ISP” mode. CRP3 disables all external access including ISP. Kinetis parts use a Flash Security mechanism with a backdoor key option. If the backdoor key was left at its default or is recoverable, firmware readout proceeds through the standard debug interface after key entry. Our engineers see roughly 60–70% feasibility on LPC1xxx/LPC2xxx parts at CRP1/CRP2; CRP3 drops that significantly.
Nordic nRF5x
Nordic’s nRF52832 and nRF52840 use the APPROTECT register to gate debug access. Earlier silicon revisions (pre-2020 errata) had a well-documented hardware vulnerability that allowed APPROTECT bypass via a carefully timed voltage glitch on the supply rail. Nordic addressed this in later revisions, making newer nRF52 and nRF53 parts substantially more resistant.
Nuvoton NUC & M-Series
Nuvoton’s Cortex-M0/M4 parts—including the NUC131 family—implement flash lock bits and a security lock register. These parts are generally among the more accessible for authorized firmware readout because the protection mechanisms rely on software-configurable registers rather than hard-blown fuses. Turnaround on Nuvoton recovery is typically 3–7 working days.
Maxim (Analog Devices) MAX326xx
The MAX32660 and related parts use a debug-lock mechanism tied to a flash information block. When locked, the SWD port responds to connection but refuses flash reads. Feasibility depends on whether the permanent lock fuse has been blown versus the software lock alone.
Cypress / Infineon PSoC
PSoC devices based on Cortex-M0/M0+ cores, such as CY8C family parts, use a protection model with chip-level protection settings (Open, Protected, Kill). The “Protected” state blocks debug but may be recoverable; “Kill” state is designed to be irreversible.
ARM Cortex-M Firmware Readout Methods
The specific technique depends on the target’s protection level, silicon revision, and available equipment. Here are the primary approaches our engineers use.
1. Standard Debug Readout (SWD / JTAG)
When protection is Level 0 or has been disabled, firmware readout is straightforward: connect a J-Link, ST-Link, or CMSIS-DAP probe, issue memory-read commands across the flash address range (typically 0x08000000 on STM32, 0x00000000 on NXP LPC), and save the resulting binary. Total time: minutes.
2. Voltage Fault Injection (Glitching)
The most widely applicable non-invasive bypass for Level 1 protection. A precisely timed voltage glitch—typically a 5–50 ns drop on the VDD or VCORE rail during the boot sequence—can cause the protection-check logic to misread the RDP/CRP bits, temporarily granting debug access. Equipment required includes a ChipWhisperer, custom FPGA glitcher, or similar pulse-injection platform, plus an oscilloscope for timing calibration. Glitch parameters vary per family: STM32F1 parts may require a glitch at ~1.2 µs after reset release, while STM32F4 timing windows shift to ~3–8 µs depending on the PLL configuration.
3. Bootloader / ISP Exploitation
Some Cortex-M parts have ROM-resident bootloaders (UART, USB-DFU, I²C, SPI) that may not fully respect flash protection under certain conditions. Known edge cases include NXP LPC parts where CRP1 allows partial ISP commands, and older STM32 parts where the system bootloader could be entered before RDP was fully enforced. These avenues are chip-revision-dependent and increasingly patched in newer silicon.
4. Invasive / Semi-Invasive Techniques
For Level 2 / permanent-lock parts, the remaining options are physical: decapping the die, using UV or laser to reset fuse bits, or direct probing of internal buses with a focused ion beam (FIB). These methods require a semiconductor lab, cost $2,000–$15,000+, and carry a non-trivial risk of destroying the die. We pursue these only when the firmware value justifies the cost and risk, and only with documented proof of ownership.
What Determines Feasibility and Turnaround?
| Factor | Impact on Readout |
|---|---|
| Protection level set | Level 0: trivial. Level 1: likely. Level 2/permanent: uncertain, often requires invasive work. |
| Silicon revision / date code | Older revisions often have known vulnerabilities. Post-errata silicon is harder. |
| External flash vs. internal | Firmware stored on external SPI flash (W25Q, AT25, SST25) can often be read directly off the bus regardless of MCU lock state. |
| Secure boot / encrypted flash | Even if readout succeeds, encrypted flash yields ciphertext. Key recovery is a separate challenge. |
| Number of sample chips available | Glitching is sometimes destructive. Having 2–3 spare MCUs improves success probability from ~70% to >90%. |
Typical turnaround for a Level 1 Cortex-M readout is 5–12 working days. Invasive work on Level 2 parts extends to 3–6 weeks depending on lab queue and decapping complexity.
Authorization and Legal Requirements

ARM Cortex-M firmware readout is a controlled service. We require documented proof that you own the IP or have authorization from the IP owner before any work begins. This includes purchase records for the hardware, design files showing your company’s origin of the product, or a signed authorization letter from the rights holder. Every engagement is covered by an NDA. For a detailed explanation of what qualifies and what we decline, see our page on whether IC unlock is legal and the jobs we turn down.
Deliverables After Successful Readout
- Raw binary file (.bin) — byte-for-byte flash image, loadable via any standard programmer.
- Intel HEX (.hex) or Motorola S-record (.srec) — address-annotated format for tools that require it.
- Option bytes / configuration dump — clock settings, brown-out levels, write-protection sectors, boot configuration.
- Readout report — documents the MCU part number, protection level encountered, method used, and verification checksum.
We verify every readout by programming the extracted binary into a blank device of the same part number and confirming functional operation before delivery.
When External Memory Makes MCU Protection Irrelevant
Many Cortex-M designs store configuration data, calibration tables, or even secondary firmware in external serial EEPROMs or SPI flash. Parts like the W25Q64, AT24C512, or M95080 can be read directly with a clip-on probe—no MCU unlock required. If your recovery goal is configuration data rather than the core application firmware, this path is faster, cheaper, and nearly 100% successful.
Frequently Asked Questions
Can ARM Cortex-M firmware readout recover source code?
No. Firmware readout extracts the compiled binary (machine code), not the original C/C++ source. The binary can be disassembled into ARM Thumb assembly, and with effort, decompiled into approximate C using tools like Ghidra or IDA Pro. But variable names, comments, and high-level structure are lost at compilation. The binary is sufficient for reprogramming identical hardware.
How much does Cortex-M firmware readout cost?
Non-invasive readout (Level 0–1 protection) typically ranges from $300–$1,500 depending on the MCU family and complexity. Invasive work on Level 2 / permanent-lock parts starts at $2,000 and can exceed $10,000 for advanced secure-boot implementations. We provide a binding quote after a free feasibility assessment.
What if my MCU uses TrustZone or encrypted flash?
TrustZone-M (ARMv8-M) and vendor-specific flash encryption (e.g., STM32H7 OTFDEC, NXP PRINCE) add a layer beyond simple readout protection. Even if we extract the flash contents, the data may be AES-encrypted. Feasibility depends on whether the encryption key is stored in a recoverable location. We assess this during the initial evaluation and give you an honest probability before you commit.
Do I need to desolder the MCU from the board?
Usually no. For non-invasive methods, we connect to the SWD/JTAG pads on the PCB. If the debug header was removed, we solder temporary wires to the MCU’s SWDIO, SWCLK, and reset pins. Desoldering is only required for invasive decapping or when the MCU must be placed in a specialized socket for fault injection.
Is the MCU destroyed during firmware readout?
For standard debug readout, the MCU is completely unaffected. For voltage glitching, there is a small risk (roughly 5–10%) of damaging the target device—this is why we recommend sending 2–3 samples when possible. Invasive decapping is inherently destructive to the package, though the die itself may survive.
Check feasibility for your project
Send the chip marking or board details and tell us what you need to recover. Add clear photos if available.
Request a feasibility reviewWhat do you need to recover?
Send the full marking or board photos, sample condition and required outcome. Ask for an NDA before sharing confidential files. Scope and sample handling are confirmed individually.
The article title is carried into the contact form. Add your own details and review the enquiry before sending.