iPhone Spyware Triage with MVT: A Practical Guide
How to check an iPhone for mercenary spyware and stalkerware: sysdiagnose, encrypted backup and file system checks with MVT, indicators, and limits.
"Is my phone hacked?" arrives at incident response teams, newsrooms and forensic labs every day. On an iPhone, the honest answer comes from a structured triage: collect the right data quickly, run it against published indicators with the Mobile Verification Toolkit (MVT), review the anomalies that no indicator covers, and report exactly what was and was not examined.
This guide covers that workflow for two different threats: mercenary spyware (Pegasus, Predator and similar exploit-based implants) and stalkerware and account abuse, which on iOS usually works through legitimate features rather than exploits. It does not explain how to develop or deploy either.
Know what you are looking for
Mercenary spyware on iOS runs as hidden processes after exploiting the device, often through zero-click attack surfaces such as iMessage and media parsing. It avoids persistence on many versions and tries to clean up after itself. What it leaves behind are small traces: process names in system databases, crash reports from failed exploit attempts, suspicious domains in browser and message data, entries in shutdown.log.
Stalkerware and coercive control on iOS rarely need an exploit. Common routes are: someone else's Apple Account signed in, or the owner's account credentials known to someone else (iCloud backups, Find My, Messages in iCloud); a configuration profile or MDM enrolment; an app with location and microphone permissions installed while the phone was unlocked; a shared family plan with location sharing. The evidence is in settings and accounts, not in exploit traces.
Decide early which scenario you are triaging. Many cases need both.
Collection order
Speed matters: unified logs keep days, crash reports weeks, and a reboot can change shutdown.log. Recommended order for a device the owner hands over unlocked:
- Document the state (screen, iOS version, battery) and keep the device powered. Do not reboot, update or reset anything yet.
- Trigger a sysdiagnose (both volume buttons and the side button briefly) and note the time. Retrieve it through Settings > Privacy & Security > Analytics & Improvements > Analytics Data, or with
pymobiledevice3. See the sysdiagnose guide. - Pull crash reports with
idevicecrashreport -korpymobiledevice3 crash pull, keeping them on the device. - Take an encrypted backup with a password you set (see acquisition).
- Full file system extraction if you have lawful access to a tool that supports the model and build: it reaches
netusage.sqlite, the liveshutdown.log, caches and more. - Screenshots of Settings > General > VPN & Device Management, the Apple Account page and its device list, and Settings > Privacy & Security > Safety Check if relevant.
Hash every archive as you collect it.
Running MVT
MVT is a command-line tool from Amnesty International's Security Lab. It extracts records from iOS backups and file system dumps into JSON, builds a timeline, and flags matches against indicator files in STIX2 format.
pip install mvt
mvt-ios download-iocs # fetch published indicator files
mvt-ios decrypt-backup -p '<backup-password>' -d ./case/decrypted ./case/backup/<UDID>
mvt-ios check-backup --output ./case/mvt-backup ./case/decrypted
mvt-ios check-fs --output ./case/mvt-fs ./case/ffs # if you have a file system dump
Indicator files published by Amnesty, Citizen Lab and others are loaded automatically after download-iocs, or passed explicitly with --iocs file.stix2. Results can be re-checked against new indicators later with mvt-ios check-iocs.
MVT writes one JSON file per module, a timeline.csv, and *_detected.json files plus timeline_detected.csv for matches. Read the console output: modules that found no database say so, which tells you what was not examined.
What the modules look at
| Source | MVT module(s) | Mode | Page |
|---|---|---|---|
| Process network usage | Datausage, Netusage | backup + fs, fs only | DataUsage and netusage |
shutdown.log | ShutdownLog | fs | shutdown.log |
| Configuration profiles and events | ConfigurationProfiles, ProfileEvents | backup | configuration profiles |
| Privacy permissions | TCC | backup + fs | TCC.db |
| Location clients | LocationdClients | backup + fs | locationd |
| Installed apps | Applications | backup + fs | installed apps |
| Messages and attachments | SMS, SMSAttachments, WhatsApp | backup + fs | sms.db, WhatsApp |
| Browsing | SafariHistory, SafariBrowserState, WebKit, Chrome and Firefox modules | backup + fs | Safari |
| Contact interactions | InteractionC | backup + fs | interactionC.db |
| Analytics and crash data | Analytics, OSAnalyticsADDaily | fs, backup + fs | crash logs |
| IDS status cache, calls, calendar, shortcuts | various | backup + fs |
Module names and coverage change between MVT releases; check mvt-ios check-backup --list-modules on your version.
Reading the output
Start with the console log and the *_detected.json files. Each detection names the module, the matched indicator and the record that matched: a URL in an SMS, a process name in DataUsage, a profile UUID. Open the underlying record in the module's full JSON output to see its context (timestamps, sender, neighbouring rows), then find the same record in the original database with sqlite3 so that the finding does not rest on the tool alone.
Next, read timeline.csv around each detection. A malicious link in a message received at 10:02, followed by crashes of a media parsing process at 10:03 and a new process name in DataUsage first seen at 10:05, tells a coherent story. A detection with nothing around it is weaker.
Then read the modules that returned nothing. A module that "found no database" did not check that source. On an unencrypted backup, several modules have nothing to read; on a backup without netusage.sqlite, the Netusage module cannot run. List those gaps in the report.
Finally, keep the raw output. Indicator sets grow after investigations are published, and a device checked today may produce a match next year against the same stored data.
Beyond indicators: manual review
Indicators only catch what someone already documented. Spend time on anomalies:
- Process names in DataUsage, netusage, JetsamEvent reports and
shutdown.logthat do not exist on a clean device of the same build. Keep a baseline extraction from a reference device. - Paths in
shutdown.logunder/private/var/db/or/private/var/tmp/, the pattern Kaspersky's research associated with Pegasus, Reign and Predator. iShutdown and iLEAPP flag them. Treat them as leads. - Crash clusters in messaging and media processes around a date of interest.
- Links in messages that point to unfamiliar domains, especially one-off domains received shortly before a crash cluster.
- Profiles that install root certificates, VPNs or proxies, and MDM enrolment by an unknown organisation.
For each lead, look for the same name, domain or time in at least one independent source before calling it significant.
Stalkerware and account abuse checks
- Accounts. Which Apple Account is signed in, and which other devices are on it (Settings > [name] > device list). Accounts3.sqlite lists every configured account with creation dates.
- Sharing. Find My location sharing, Family Sharing, shared albums and shared Notes. Safety Check (iOS 16+) summarises sharing and access and can reset it; do not run the reset before documenting.
- Profiles and MDM. Configuration profiles, with install events.
- Apps and permissions. Installed apps not from the App Store, and TCC and locationd grants for camera, microphone and location.
- Backups and computers. An iCloud backup accessible to another person exposes almost everything. A trusted computer that is not the owner's can pair and back up the device.
Coordinate with the person at risk before removing anything: removing access can alert an abuser. Organisations such as the Access Now Digital Security Helpline and local domestic violence services can help plan that step.
Interpreting results
| Outcome | What you can say |
|---|---|
| Indicator match in several independent sources, consistent timing | Strong evidence of a known compromise; escalate to a specialist lab and preserve everything |
| Single indicator match | A lead requiring confirmation; check for false positives (for example, a shared domain) |
| Anomalies without indicator matches | Unexplained findings; describe them, compare with a baseline, seek expert review |
| Nothing found | No known indicators in the data examined; list the sources examined and not examined |
Never tell someone their phone is "clean". Say what was checked.
Baselines and false positives
Indicator matching is only as good as the indicators. Domains used by spyware operators are sometimes reused or re-registered, and some indicator sets include process names that are generic enough to collide with legitimate software. Before escalating a single match, check whether the indicator is specific (a unique domain seen in a message link, a process name with no legitimate use) or generic, and whether the timing makes sense.
For anomalies without indicators, a baseline is the best defence against false alarms. Keep reference data from a clean device of the same model family on the same iOS build: a sysdiagnose, a backup and, if possible, a file system extraction. Compare process names, paths in shutdown.log, DataUsage process lists and crash report types. Many "suspicious" daemons turn out to be new Apple services introduced in a point release.
Documenting the triage
A triage report should let another analyst repeat the work. Record the device (model, build, state), each collection with time, method and hash, the MVT version and the indicator files used (names, dates, hashes), the modules that ran and those that found no data, every lead with its sources, and the conclusion in the terms of the table above. Keep MVT's JSON output and timelines with the case: when new indicators are published, mvt-ios check-iocs can re-examine them without touching the device again.
After triage
If compromise is confirmed or strongly suspected, preserve the device and the collected data, and contact an experienced lab (Amnesty Security Lab and Citizen Lab work with civil society targets). For at-risk users, Lockdown Mode (iOS 16+) reduces attack surface significantly, and updating to the latest iOS removes known exploit paths. Updating and rebooting destroy volatile evidence, so collect first.
Common mistakes
- Rebooting or updating the device before collecting a sysdiagnose.
- Running MVT on an unencrypted backup and missing sources only encrypted backups contain.
- Reporting "no spyware" instead of "no known indicators found".
- Treating a
/private/var/db/path inshutdown.logas proof without corroboration. - Ignoring account and profile abuse because MVT reported no detections.
- Removing a suspicious profile or account before documenting it and discussing safety with the person.