Wearable forensics: Apple Watch, Wear OS and fitness trackers
Where smartwatch and fitness-tracker data sits (watch, phone, cloud, exports), what Apple and Google document, dated vendor claims, and the gaps.
Updated
In short
Wearable data can sit in up to four places: on the watch, on the paired phone, in a vendor cloud, and in exported files, and which holds the most depends on the device (one two-watch study found more on the watch for one and on the phone for the other). Apple documents that an Apple Watch is backed up to its paired iPhone, or to the family member's iCloud account for a family-managed watch, and Elcomsoft and MOBILedit state that they can extract directly from Apple Watch models, vendor claims with no independent test found. For Wear OS the directory has no tool built for it, only small unvalidated GitHub projects were found, MOBILedit's manual states Wear OS support, and the vendor-stated Galaxy Watch features are partly chipset-specific. Two NIST CFTT reports recorded fitness data from wearable companion apps as not reported on the phones tested, alongside other gaps, though MD-LIVE reported Apple Health data; of the six studies reviewed, four are single-model case studies and five were read by abstract only.
Where wearable data lives
SWGDE 18-F-003-2.0 (2025-08-21) tells examiners to search the scene for paired or linked devices, naming smartwatches, and says data related to a mobile device can often be found on a smartwatch because of synchronisation through a backup process or a cloud service account.
- The watch. The studies below found messages, email and other data on the watch itself, on specific models.
- The paired phone. Synced data, including the companion app's records.
- The vendor cloud. Google's Data Layer documentation says Data Layer traffic between a watch app and its phone app can travel by Bluetooth or through Google's servers as an intermediary, and that data routed through Google Cloud is end-to-end encrypted.
- Exported files. FIT, GPX and SQLite files exported by Gadgetbridge, and CSV and JSON files in Samsung Health data (see the Marjianto abstract below).
One study gave different answers for two devices. The abstract of Odom et al. (2019) (abstract read, not the full paper) reports significant amounts of data on a Samsung Gear S3 Frontier, greater than on its companion phone, but for an Apple Watch Series 3 it found that the companion phone stored the greatest amount of data.
Apple Watch
Pairing and backup. Apple's watchOS security chapter (published 2026-01-28) says an Apple Watch pairs with only one iPhone at a time and that, when it is unpaired, the iPhone sends instructions to erase all content and settings from the watch. Apple's restore guide says Apple Watch content is backed up continuously to the paired iPhone, that a backup is performed first if the devices are unpaired, and that Apple Watch backups are included when the iPhone is backed up to iCloud or to a computer. The same page says a watch managed for a family member backs up directly to the family member's iCloud account when it is connected to power and Wi-Fi. This guide reads the unpairing rule as a reason to treat unpairing as destructive for the watch's own data.
Health data. Apple's health data chapter (published 2026-01-28) says HealthKit holds health and fitness data from iPhone and Apple Watch, stored in the Data Protection class Protected Unless Open, with access relinquished 10 minutes after the device locks. Health data in iCloud is end-to-end encrypted when the account uses iOS 12 or later and two-factor authentication. A Finder or iTunes backup stores health data only if it is encrypted, and Apple's backup page (published 2026-09-21) lists health data among the items an encrypted backup can include. The class is different from the AFU-readable class described in BFU vs AFU; none of the sources reviewed say what that means for any particular extraction method.
What published examinations found. Alabdulsalam et al. (arXiv, 2018) examined an Apple Watch Series 2 (the paper's specification lists watchOS 2.3, 3.0 and an upgrade to 3.2; the version examined is not stated). The authors reported a hidden diagnostic port with no official cable, so they used a Cellebrite UFED logical extraction of the iPhone plus manual browsing of the watch. On the iPhone they found the paired watch in a Bluetooth database and as a source device in the Health database. With the iPhone off, messages, email, contacts, pictures and voicemail were still readable on the watch. The authors state that a PIN-locked watch could not be examined by the manual method without the PIN. This is one early model, not a statement about current watches.
Vendor claims. As of Elcomsoft's iOS Forensic Toolkit page retrieved on 2026-11-10, the vendor states that it supports all Apple Watch generations: bootloader-level extraction for Series 0 to 5, logical extraction for Series 0 to 10, SE, SE2 and Ultra models, with a third-party adapter needed to connect. See the Elcomsoft iOS Forensic Toolkit entry. Reincubate's product page (retrieved 2026-11-10) lists "Apple Watch data recovery" in the paid edition; per its entry, Reincubate iPhone Backup Extractor is aimed at people recovering their own data and has no forensic workflow. MOBILedit's manual page for Apple Watch states that MOBILedit Forensic extracts from Apple Watch Series 0 to 6 and SE models through an All-in-One Reader and from Series 7 to 12 and Ultra models through a wireless adapter, that the PIN must be known, and that the wireless reader needs a driver due in version 9.9. The manual's physical-port page lists only Series 0 to 5 and SE 1st and 2nd generation. No independent test of any of these claims was found.
Wear OS
What Google documents. The Data Layer page (last updated 2026-11-09) says the API is available only on Wear OS watches and paired Android devices. The standalone apps page (last updated 2026-10-23) defines a standalone watch app as one that does not require a phone app for core features, and says watch apps can store data locally with standard Android storage APIs such as DataStore or Room. None of the Android developer pages reviewed describe where health data is stored on the watch or how an examiner would acquire it.
Published research. The abstract of Marjianto and Luthfi (2025) (abstract read, not the full paper) describes a Samsung Galaxy Watch 5 paired with a Samsung S22. It lists file system exploration over Android Debug Bridge (ADB), physical forensics and data collection, and reports heart rate, sleep, step, calorie and exercise data from the Samsung Health application as CSV and JSON files. The abstract does not say whether those files came from the watch or the phone.
Vendor features. The vendor items cited here are for Samsung Galaxy Watch models. Some are chipset-specific and some are not described by chipset in the notes cited.
- A Forensic Focus news post of 2026-10-01 states that Passware Kit Mobile 2026 v5 recovers passcodes and extracts data from 24 Galaxy Watch 4, 5 and 6 models built on the Exynos W920 or W930 with eMMC storage.
- MOBILedit's 9.7 notes (2025-11-05) say Exynos bypassing now covers Galaxy Watch 5, 6 and 7, previously only Watch 4, in the Ultra edition. The 9.8.1 notes (2026-08-05) add analysis of Samsung Watch sensor data. See MOBILedit Forensic.
- MOBILedit's undated Smartwatch Kit page says its Samsung Galaxy Watch readers can obtain messages, geolocations, health data and heartbeats from the Galaxy Watch 2 up to the latest versions; the manual's Samsung Watch page says earlier Samsung watches run Tizen and later ones Wear OS, and that Samsung Health data needs the Ultra edition with Exynos decrypt.
Neither the Passware nor the MOBILedit news item describes a general Wear OS method, and their model lists differ. MOBILedit's undated manual says Wear OS is used by Samsung, Suunto and other brands, that it supports extraction from Huawei smartwatches running WearOS, and that it should support extractions from other Wear OS smartwatches. These are vendor statements and no independent test was found.
Dedicated tools. This directory has no tool built for Wear OS. ALEX, an ADB script, is the only directory entry that names WearOS (README, tag v0.7: "extractions are possible for Android, FireOS, WearOS and Ubuntu Touch devices"); the README describes no WearOS-specific features. The only dedicated projects found in public sources are two small GitHub repositories: ITeratOr-WearOS (one star, no tags, last pushed on 2024-12-03), which describes itself as a WearOS forensics project, and IoTrace (two stars, no tags, last pushed on 2025-03-11), which describes itself as a non-invasive forensic tool for WearOS smartwatches. No published validation of either was found, this guide did not test them, and neither has a directory entry. For phone lock-state behaviour see BFU vs AFU; no source reviewed describes the equivalent for watches.
Garmin, FIT files and fitness trackers
The FIT format. Garmin's FIT overview describes FIT as a protocol for storing and sharing data from sport, fitness and health devices. The Garmin FIT SDK is available in eight languages, including Python. At tag 21.221.0 (the newest tag on 2026-11-10) the Python README describes an integrity check of header string, file size and CRC and decoding options that are all enabled by default, including scale/offset and datetime conversion.
Open-source parsers. python-fitparse parses FIT files and includes a --ignore-crc option for devices that write invalid CRCs. The default-branch README (the README at the newest tag, v1.2.0, has no such notice) says the maintainer has limited to no time, is seeking a maintainer and suggests the fitdecode library as an alternative. GarminDB (tag v3.9.0) imports Garmin data into SQLite, but the README documents downloading from Garmin Connect with the account holder's username and password in a configuration file. Use of account credentials is a legal question; see cloud extraction, tokens, backups and legal process.
Gadgetbridge. Gadgetbridge is an open-source Android app that, per its tool entry, replaces the vendor's own app for supported wearables. Its data management documentation warns that uninstalling or reinstalling without an export loses the data. Its auto-export documentation says that for devices providing their own raw .fit file (Garmin, iGPSPORT) the original is exported unchanged, while for others a FIT file is synthesised from parsed workout data, a list that includes Xiaomi, Garmin, Huami/Amazfit, Huawei, CMF and Bangle.js devices. Whether a Gadgetbridge FIT file is the device's original therefore depends on the device.
Other trackers. The abstract of Yoon and Karabiyik (2020) (abstract only) reports that the Fitbit Versa 2 companion app on Android did not store pushed social media notifications, but that some credit card information, heart rate and GPS data were found in plaintext. The abstract of Dawson and Akinbi (2021) (abstract only) reports artefacts from a TomTom Spark 3 watch, its Android app and Bluetooth event logs showing when activity was synced and uploaded to the cloud.
Limits
- Parsing gaps in phone tools. The CFTT report on Data Pilot Desktop v2.0.200 (December 2024) lists fitness data as not reported for phones paired with, among others, Fitbit, Garmin, Coros and Galaxy Watch 6 devices. The report on GMDSOFT MD-LIVE v3.6.14.1440 (February 2026) lists a similar result for seven wearables, among them the Galaxy Watch 6 and iWatch. See the Data Pilot Desktop and MD-LIVE entries. The Data Pilot tables also mark GPS, Wi-Fi and social-media data as not as expected on phones tested. MD-LIVE says fitness data was populated onto its iOS devices using the wearables (section 4.2); Data Pilot's population objects include fitness-application data (section 3.2). MD-LIVE reports that Apple Health data was reported on the iPhones, but not the data held in the Garmin, iWatch, Coros and Fitbit companion apps.
- Sync gaps. In the Apple Watch Series 2 study, the messages and pictures readable on the watch with the iPhone off were those synchronised before it was turned off. This guide infers that data not yet synced to a phone or cloud may exist only on the watch.
- Cloud-only and locked data. Google says Data Layer traffic may route through its servers and that data transferred through Google Cloud is end-to-end encrypted, and Apple says Health data in iCloud is end-to-end encrypted under stated conditions. The passcode-related claims above from the Passware and MOBILedit news items cover specific Galaxy Watch models only, and the Elcomsoft page reviewed does not describe passcode-locked watches.
- Validity of the data itself. The abstract of Sicbaldi et al. (2026) (abstract only) reports that validation evidence for smartwatch measurements is heterogeneous, with heart rate the most studied and most accurate measure. This guide therefore treats step counts, sleep stages and calorie values as device estimates until validated. See validating mobile forensics tools before relying on any vendor claim here.
Sources
- SWGDE 18-F-003-2.0, Best Practices for Mobile Device Evidence Collection & Preservation Handling and Acquisition (Version 2.0, 2025-08-21)
- Apple Platform Security Guide, System security for watchOS (published 2026-01-28)
- Apple Platform Security Guide, Protecting access to user's health data (published 2026-01-28)
- Apple Watch User Guide, Restore Apple Watch from a backup
- Apple Support, About encrypted backups on your iPhone, iPad, or iPod touch (published 2026-09-21)
- Alabdulsalam, Schaefer, Kechadi, Le-Khac, Internet of things forensics: Challenges and Case Study (arXiv, January 2018; Apple Watch Series 2 case study)
- Odom, Lindmar, Hirt, Brunty, Forensic Inspection of Sensitive User Data and Artifacts from Smartwatch Wearable Devices, Journal of Forensic Sciences (2019); abstract only
- Elcomsoft iOS Forensic Toolkit product page (vendor), retrieved 2026-11-10
- Reincubate iPhone Backup Extractor product page (vendor), retrieved 2026-11-10
- Android Developers, Overview of Data Layer API (Wear OS), page last updated 2026-11-09
- Android Developers, Standalone versus non-standalone Wear OS apps, page last updated 2026-10-23
- Marjianto, Luthfi, Investigation of Digital Evidence on Android-Based WearOS Smartwatches, ICITDA (2025); abstract only
- Forensic Focus, Passware Kit Mobile 2026 v5 Decrypts Samsung Galaxy Watch (2026-10-01; news post on a vendor release)
- MOBILedit, New release: MOBILedit Forensic 9.7 (2025-11-05, vendor)
- MOBILedit, MOBILedit Forensic 9.8.1 released (2026-08-05, vendor)
- MOBILedit, Smartwatch Kit product page (undated, vendor)
- MOBILedit Forensic manual, Apple Watch (undated, vendor)
- MOBILedit Forensic manual, Apple Watch with physical diagnostic port (undated, vendor)
- MOBILedit Forensic manual, Samsung Watch (undated, vendor)
- MOBILedit Forensic manual, WearOS applications (undated, vendor)
- MOBILedit Forensic manual, Other Smartwatch Brands (undated, vendor)
- Pranavh-2004/IoTrace on GitHub (last push 2025-03-11; no tags)
- prosch88/ALEX README (tag v0.7)
- jonnnnn-c/ITeratOr-WearOS on GitHub (last push 2024-12-03; no tags)
- Garmin Developers, FIT SDK overview
- garmin/fit-python-sdk README (tag 21.221.0)
- dtcooper/python-fitparse README (master and tag v1.2.0)
- tcgoetz/GarminDB README (tag v3.9.0)
- Gadgetbridge documentation, Auto export
- Gadgetbridge documentation, Data management
- Yoon, Karabiyik, Forensic Analysis of Fitbit Versa 2 Data on Android, Electronics (2020); abstract only
- Dawson, Akinbi, Challenges and opportunities for wearable IoT forensics: TomTom Spark 3 as a case study, Forensic Science International: Reports (2021); abstract only
- Sicbaldi et al., Is my smartwatch a valid witness? A systematic review and meta-analysis, Forensic Science International (2026); abstract only
- DHS S&T / NIST CFTT, Test Results for Mobile Device Acquisition Tool: Data Pilot Desktop v2.0.200 (December 2024)
- DHS S&T / NIST CFTT, Test Results for Mobile Device Acquisition Tool: GMDSOFT MD-LIVE v3.6.14.1440 (February 2026)