Skip to content

Mobile Device Hardware Architecture

Modern smartphones and tablets pack multiple specialised chips onto a single board, and the design choices made by manufacturers directly determine what data survives, where it lives, and how forensic tools can reach it. This topic maps the key hardware components of a mobile device and explains the forensic implications of each.

Last updated:

Share

Mobile device hardware architecture describes the physical components built into smartphones and tablets: the System on Chip that runs the main operating system, the NAND flash memory chips that store user data, the baseband processor that manages radio communications, and the secure hardware that protects encryption keys. Each component shapes what data exists on the device, where that data lives, and whether a forensic examiner can reach it. Understanding the hardware layer is the prerequisite for choosing an acquisition strategy, because the same logical acquisition tool that works on one chipset may fail entirely on another.

The dominant SoC families in current devices include Apple's A-series and M-series chips, Qualcomm Snapdragon, MediaTek Dimensity, Samsung Exynos, and Google Tensor. Each integrates CPU cores, a GPU, an image signal processor, a neural engine, and a memory controller, plus varying implementations of a secure enclave or Trusted Execution Environment (TEE). Storage is provided by separate NAND flash chips packaged as eMMC or UFS modules, soldered directly to the board. The baseband processor, which may be a discrete chip or integrated into the SoC, runs its own real-time operating system and non-volatile memory bank.

From a forensic standpoint, the most consequential hardware feature of any modern mobile device is hardware-backed encryption. The encryption key is derived from a combination of the user passcode and a device-unique key that is fused into the SoC at manufacture and never exposed outside the secure hardware boundary. This means that a chip removed from a locked device yields encrypted storage that cannot be decrypted without either the correct passcode or a vulnerability in the key derivation or key storage implementation. Hardware architecture choices made by the manufacturer set the ceiling on what any acquisition method can achieve.

By the end of this topic you will be able to:

  • Identify the main hardware components of a smartphone and explain the forensic significance of each.
  • Distinguish between eMMC and UFS storage architectures and explain how each affects chip-off acquisition.
  • Explain how hardware-backed encryption limits physical acquisition on locked modern devices.
  • Describe what data the baseband processor holds and what acquisition technique is needed to reach it.
  • Explain how NAND wear levelling produces forensically recoverable artefacts from logically deleted data.
Key terms
System on Chip (SoC)
A single integrated circuit that combines the application processor, GPU, memory controller, image signal processor, and often the secure enclave. The SoC model determines the encryption scheme, secure boot chain, and available hardware debugging interfaces.
eMMC (embedded MultiMediaCard)
A flash storage standard that packages NAND memory chips and a controller into one soldered module using a parallel interface. Common in mid-range and older devices. Its single-die structure makes chip-off acquisition relatively straightforward with widely available readers.
UFS (Universal Flash Storage)
A newer flash storage standard using a serial interface that allows simultaneous reads and writes. Faster than eMMC and common in flagship devices from 2018 onward. Chip-off acquisition requires specialised UFS readers and is more complex than eMMC.
Baseband processor
A separate processor that manages all radio functions: cellular calls, SMS, and data connections. It runs its own real-time OS and holds its own non-volatile memory with call routing metadata and SIM interaction logs. Not reachable through standard logical acquisition.
Secure Enclave / Trusted Execution Environment (TEE)
A hardware-isolated execution environment within the SoC that stores device-unique encryption keys and handles cryptographic operations. The encryption key never leaves this boundary, making it the primary barrier to decryption on locked modern devices.
Flash Translation Layer (FTL)
Firmware inside the flash storage controller that maps logical block addresses to physical NAND blocks and implements wear levelling. Deleted files may persist in out-of-mapping physical blocks until the FTL recycles them, creating a window for forensic recovery.

The System on Chip: architecture and forensic significance

The SoC is the central decision-making component. Apple's A-series chips (A14 onward) and Google's Tensor G-series tightly integrate the Secure Enclave or Titan M security chip directly onto the same package, giving a small physical attack surface. Qualcomm Snapdragon SoCs implement the TEE through ARM TrustZone software partitioning, with the Secure Processing Unit handling key storage. MediaTek Dimensity SoCs follow a similar TrustZone model. These architectural differences mean that the same exploit that bypasses the TEE on one SoC generation does not transfer to another.

The secure boot chain matters forensically because it determines whether custom recovery environments, bootloaders, or debugging modes can be activated. On locked Apple devices, the Secure Boot process verifies each stage from the Boot ROM upward using Apple-signed keys. An unsigned image will not load. On Android devices with an unlocked bootloader, a forensic boot image can be loaded to extract a full file system image without touching device storage. With a locked bootloader, that path is closed and the examiner must rely on logical extraction through the OS or on hardware techniques.

The application processor cores within the SoC run the mobile OS and all user applications. They typically use ARM Cortex-A cores in a big.LITTLE configuration: high-performance cores for demanding tasks and efficiency cores for background work. RAM is provided by LPDDR4X or LPDDR5 chips, packaged separately or in a Package on Package (PoP) stack directly above the SoC. RAM is volatile and loses its contents seconds to minutes after power loss, but in live acquisition scenarios or when the device is found powered on, RAM may contain decrypted keys, active session tokens, and plaintext message content that is not available in storage.

NAND flash storage: eMMC, UFS, and forensic recovery

NAND flash is the non-volatile storage technology in every modern smartphone. It stores the OS, applications, and all user data. NAND cells are organised into pages (the minimum read/write unit, typically 4 to 16 KB) and blocks (the minimum erase unit, typically 256 to 512 pages). Writing to NAND requires an erase-before-write cycle at the block level, which creates the need for the Flash Translation Layer.

FeatureeMMCUFS
InterfaceParallel (MMC)Serial (MIPI M-PHY)
Simultaneous read/writeNoYes
Typical max speed400 MB/s2100 MB/s (UFS 3.1)
Common device eraPre-2018 mid/low range2018 onward flagships
Chip-off adapter availabilityWideLimited, specialist hardware
Package identificationBGA-153 or BGA-169 commonBGA-254 or embedded

For chip-off acquisition, the storage chip is removed from the board using hot-air rework or infrared soldering equipment, then read directly by a chip reader that bypasses the device OS entirely. eMMC chips are well-supported by forensic chip readers from vendors such as Cellebrite, MSAB, and Octoplus. UFS chips require dedicated UFS socket adapters and more specialised equipment. The extracted image is a raw NAND dump, which must then be parsed by software that understands the file system structure and the FTL's address mapping.

Wear levelling creates the main forensic opportunity in NAND flash. When the OS deletes a file, it updates the file system metadata to mark those blocks as free. The FTL does not immediately erase the underlying NAND blocks; it marks them as invalid and reclaims them for new writes when it needs the space. Until that recycling happens, the original data remains in the physical NAND cells. Physical acquisition tools that read raw NAND can find these out-of-mapping blocks and recover data that the logical file system reports as deleted. The window of recovery is unpredictable: on a lightly used device with plenty of free space, deleted files may persist for weeks; on a heavily used device, the FTL may recycle blocks within hours.

The baseband processor and radio subsystem

The baseband processor, sometimes called the radio processor or modem chip, manages all wireless communication between the device and cellular networks. On many Android devices it is a discrete chip: Qualcomm MDM, Intel XMM, or Samsung Shannon series are common examples. On Apple devices since the iPhone 12, the modem is integrated into the same package as the main SoC, though it still runs its own firmware and maintains its own data partition.

The baseband processor runs its own real-time operating system, separate from Android or iOS. It has its own non-volatile memory, typically a small NOR flash or dedicated NAND partition, that stores cellular network parameters, IMEI, band preferences, and logs of network events. These logs can include cell tower registration records, handover events, and SIM interaction history. That data is forensically relevant in call record investigations, location analysis, and IMEI tracking cases.

The radio subsystem also includes a separate GPS module or integrated GNSS block, a Wi-Fi and Bluetooth combo chip, and an NFC controller. Each of these components may maintain its own small cache of recent activity. GPS modules may hold recent fix data; Wi-Fi chips may hold a list of recently connected SSIDs and their authentication parameters. On most devices these caches are managed through the application processor and appear in the main file system, but on some implementations they are stored in the peripheral chip itself.

Hardware-backed encryption and its forensic implications

Every current flagship smartphone uses hardware-backed encryption. The encryption key hierarchy starts with a device-unique hardware root key, sometimes called the UID key on Apple devices or the Hardware Derived Key on Qualcomm platforms. This key is fused into the SoC at manufacture and is not readable by any software running on the device, including the OS itself. Cryptographic operations that use this key are performed inside the Secure Enclave or TEE; the key never appears in the register file or RAM of the application processor.

iOS uses a two-factor key derivation: the hardware UID key is combined with the user passcode using a key derivation function (PBKDF2 with many iterations on modern devices) to produce the per-file encryption key. If you have the passcode, you can derive the key on the device. Without the passcode, the hardware key is inaccessible and the storage image cannot be decrypted offline. The practical effect is that a chip-off extraction of a locked, current-generation iPhone yields an encrypted image that cannot be cracked with off-device compute power, regardless of how much processing time is available.

Android devices with Android 9 or later and modern hardware use file-based encryption (FBE) rather than full-disk encryption (FDE). FBE assigns different encryption keys to different categories of data. Some files, such as the alarm clock settings, are encrypted with a key available before first unlock (BFU state). Others, including most user data, are encrypted with a key that is only derived after the correct passcode is entered (AFU state). Forensic tools such as Cellebrite UFED and Oxygen Forensic Detective attempt to extract data in the AFU state when they can obtain a file system image, because in AFU the keys are loaded in memory and the OS exposes files through its normal interfaces.

The forensic significance of the BFU versus AFU distinction is large. A device found powered off and seized is in BFU state once powered on; most user data is inaccessible through any logical means. A device found powered on and unlocked is in AFU state; a full file system image may be extractable immediately. A device found powered on but screen-locked may be in AFU state if the user had unlocked it since last reboot; tools that exploit the OS through the USB interface can sometimes access AFU-state data without the passcode. This is why first-responders are instructed to keep the device powered on, isolated from networks, and avoid letting it lock if possible.

Hardware debugging interfaces: JTAG, EDL, and chip-off

Mobile devices expose hardware debugging interfaces that were intended for factory testing and repair but are used by forensic examiners when software-based acquisition is not possible. The three main interfaces are JTAG, Qualcomm's Emergency Download Mode (EDL), and MediaTek's Preloader/Download mode. Each provides a different level of access and carries different risks.

JTAG (Joint Test Action Group, IEEE 1149.1) is a standard boundary scan interface present on most mobile SoCs. It allows a connected debugger to halt the processor, read and write memory, and access internal registers. For forensics, JTAG is used to dump the contents of flash storage by reading through the memory controller. The technique requires soldering fine wires to test points on the board, which is complex and carries a risk of permanent damage. JTAG acquisition is described in detail in the JTAG and Chip-Off Acquisition topic.

Qualcomm EDL mode (Emergency Download Mode, also called 9008 mode for its USB product ID) is a low-level boot mode built into the Qualcomm boot ROM. When triggered by a button combination or by grounding a test point, the device presents as a Qualcomm diagnostic interface over USB. With the correct Firehose programmer binary for the specific SoC, this interface can read full flash storage dumps without loading the main OS. EDL mode was the basis for many forensic tools on Qualcomm devices before vendors began restricting the Firehose programmers. Some devices have PKM (Programmer Key Manifest) restrictions that require a signed programmer, limiting who can use this technique.

TechniqueHardware contact requiredRisk levelData completeness
Logical acquisitionNo (USB/wireless)Very lowFile system (AFU state only)
File system acquisitionNo (USB)LowMore complete than logical
EDL / Download modeSometimes (test point)MediumFull flash dump
JTAGYes (soldered wires)HighFull flash dump including baseband area
Chip-offYes (board disassembly)Very high (destructive)Raw NAND including wear-levelled blocks

Hardware design variation across manufacturers and its forensic impact

Apple, Samsung, Google, Xiaomi, and other manufacturers all make different choices about how to implement the hardware components described above, and those choices create real differences in forensic outcomes. Apple's tight integration of the Secure Enclave with the application processor means that the hardware key is never accessible outside the chip boundary, and Apple has consistently resisted creating backdoors in this design. The result is that locked, fully updated iPhones present the highest barrier to physical extraction of any mainstream device class.

Samsung's Galaxy devices use Qualcomm Snapdragon or Samsung Exynos SoCs depending on the market region. Exynos devices have historically had fewer publicly known forensic exploit paths than Snapdragon devices, partly because the Qualcomm EDL ecosystem is more mature. Samsung also implements Knox, a hardware-backed security architecture that includes a physical fuse that permanently records whether the bootloader has ever been unlocked. A Knox-blown device cannot restore its attestation status, which affects both data security and the interpretation of forensic evidence: if Knox is blown, the device's security state has been altered.

From a legal standpoint, the hardware-determined encryption barrier has produced significant case law and legislative debate. In the US, the Department of Justice has sought court orders compelling Apple and Google to assist with device decryption, leading to public disagreements about the limits of lawful assistance. In the UK, the Investigatory Powers Act 2016 permits notices requiring technical assistance but does not explicitly compel the creation of decryption backdoors. In India, the Information Technology Act 2000 and the Digital Personal Data Protection Act 2023 address data access obligations, but hardware-level encryption remains outside the reach of any purely legal demand. EU member states operate under the European Investigation Order framework for cross-border requests.

Understanding which SoC, storage type, and security implementation a target device uses is the starting point for any physical acquisition strategy. See Physical Acquisition Techniques for how these hardware characteristics translate into specific acquisition decisions, and Data Persistence and Evidence Locations for how data survives at the storage layer.

Check your understanding
Question 1 of 4· 0 answered

A forensic examiner receives a locked iPhone with a current iOS version. They perform a chip-off extraction and obtain a raw NAND dump. What is the likely outcome?

Key Takeaways

  • The SoC model determines the encryption scheme, secure boot chain, and available hardware debugging interfaces. Identify the SoC before selecting an acquisition strategy.
  • eMMC storage is broadly supported for chip-off acquisition; UFS requires specialist hardware. The storage type is a practical constraint on which tools can be used.
  • Hardware-backed encryption ties the decryption key to the user passcode and a device-unique hardware key that never leaves the Secure Enclave or TEE. A chip-off dump of a locked modern device yields encrypted data that cannot be decrypted offline.
  • The BFU versus AFU state of a seized device determines what data is accessible through logical acquisition. Keeping the device powered on in AFU state is the single most consequential first-responder decision in mobile forensics.
  • NAND wear levelling means logically deleted files may remain in out-of-mapping physical blocks. Physical acquisition or chip-off can recover this data; the recovery window depends on device usage and available free space.
What is a System on Chip and why does it matter for mobile forensics?
A System on Chip (SoC) integrates the CPU, GPU, memory controller, image signal processor, and often the secure enclave onto a single die. For forensics, the SoC model determines the encryption scheme, the secure boot chain, and which hardware debugging interfaces are exposed. Knowing the SoC in a target device tells you which acquisition tools and exploit paths are plausible before you touch the device.
What is the difference between eMMC and UFS storage and how does each affect acquisition?
eMMC (embedded MultiMediaCard) uses a single-lane parallel interface and packages the NAND flash with a controller on one chip. UFS (Universal Flash Storage) uses a serial interface, supports simultaneous reads and writes, and is typically faster. For chip-off acquisition, eMMC chips are easier to read directly because adapters and readers are widely available. UFS chips require more specialised hardware and the interface differences mean some older forensic readers cannot handle them.
What does the baseband processor do and what data does it hold?
The baseband processor manages all radio communications: cellular calls, SMS routing, and data connections. It runs its own operating system, separate from the main application processor. Its non-volatile memory can hold call routing metadata, network registration records, and SIM interaction logs. Baseband memory is not accessible through standard logical acquisition; specialised hardware or JTAG techniques are required.
How does hardware-based encryption affect forensic data recovery on mobile devices?
Modern SoCs implement full-disk or file-based encryption using keys derived from the user passcode and a device-unique hardware key stored in the Secure Enclave or Trusted Execution Environment. The hardware key never leaves the secure hardware boundary, so even a chip-off extraction produces encrypted data that cannot be decrypted without the passcode or a known vulnerability. This is the primary barrier to physical acquisition on locked, modern devices.
What is NAND wear levelling and why does it create forensic artefacts?
NAND flash cells degrade after a fixed number of write cycles. The flash translation layer in the storage controller implements wear levelling by distributing writes across all available blocks rather than repeatedly writing to the same location. As a by-product, old data blocks are not immediately overwritten; they remain marked as invalid until the controller recycles them. This means logically deleted files can persist in these out-of-mapping blocks and may be recoverable through physical acquisition or chip-off analysis.

Test yourself on Mobile and Network Forensics with free, timed mocks.

Practice Mobile and Network Forensics questions

Found this useful? Pass it along.

Share

Spotted an error in this page? Report a correction or read our editorial standards.

Your journey to becoming a forensic professional starts here.

Practice with mock tests, learn from structured notes, and get your questions answered by a global forensic community, all in one place.