Skip to content

iOS Pattern of Life: KnowledgeC, Biome and PowerLog

How to reconstruct iPhone use from knowledgeC.db, Biome streams, PowerLog, Screen Time and network usage: sources, overlaps, validation and reporting.

Published on 9 min read

A messages database tells you what was said. Pattern-of-life data tells you what the phone was doing: screen on at 02:14, Instagram in the foreground for eleven minutes, locked at 02:26, plugged in at 02:31, connected to the home Wi-Fi at 02:32. For alibi checks, driving cases, workplace incidents and any question of "was the phone in use at that moment", these sources decide the case.

iOS spreads this data across several stores that overlap, disagree in small ways and change with every release. This guide explains what each one records, how to combine them, how to validate what you see, and how to report it without claiming more than the data supports.

The sources

SourceGranularityMain contentTypical retentionPage
knowledgeC.dbIntervalsApp focus (older releases), lock, backlight, plug-in, media, SiriAbout four weeksknowledgeC and Biome
Biome streamsEventsApp focus, backlight, lock, battery, Wi-Fi, Bluetooth, Safari, notificationsDays to weeks per streamBiome streams
PowerLogSamples and aggregatesApp runtime, screen state, camera, torch, battery, audioDays, weeks with archivesPowerLog
Screen TimeHourly bucketsMinutes per app and domain, pickups, per device and userAbout four weeksScreen Time
DataUsage / netusagePer processNetwork use, first and last seenMonthsDataUsage and netusage
interactionC.dbEventsContact interactions per appWeeks to monthsinteractionC.db

All but interactionC and DataUsage need a full file system extraction. Most are protected by data protection classes that require the device to be in AFU state. A sysdiagnose adds recent PowerLog copies and unified logs.

knowledgeC.db and Biome: the core

knowledgeC.db, written by CoreDuet, stores intervals in a single ZOBJECT table, one stream per ZSTREAMNAME (/app/inFocus, /device/isLocked, /display/isBacklit, /device/isPluggedIn...). For years it was the centre of iOS pattern-of-life work, and APOLLO's knowledgeC modules are still the reference queries.

From around iOS 16, Biome became the primary recorder for many of the same signals. Biome writes each stream to its own folder of SEGB files under /private/var/mobile/Library/Biome/streams/ and /private/var/db/biome/streams/. App.InFocus holds app focus; Device.Display.Backlight, _DKEvent.Keybag.IsLocked, _DKEvent.Device.IsPluggedIn and _DKEvent.Device.BatteryPercentage hold device state; _DKEvent.Wifi.Connection and Device.Wireless.Bluetooth hold connections; Safari.Navigations, Media.NowPlaying and Notification.Usage add context.

Practical rules:

  • Examine both. On current releases, knowledgeC can be sparse while Biome is rich; older events may survive only in knowledgeC.
  • Separate local and remote. Biome remote/<device> folders contain events synced from the owner's other devices. They are not events of this iPhone. knowledgeC has similar sync-related sources.
  • Collect WAL files and tombstones. Recent knowledgeC rows often live only in the WAL. Biome tombstone folders and deleted-state records can hold data the current files do not.

PowerLog: the independent witness

PowerLog was built for battery accounting, not for pattern of life, which is what makes it valuable: it records similar signals through a completely different code path. PLAppTimeService_Aggregate_AppRunTime gives foreground and background runtime per app per window; screen state tables give the frontmost app at each display change; camera, torch and audio tables show hardware use.

Two cautions. First, PowerLog timestamps are Unix seconds on an internal clock that can drift far from wall time. Apply the PLStorageOperator_EventForward_TimeOffset correction to every row, and report the offset you applied. Second, sysdiagnoses and rotated archives contain overlapping copies of the same periods; deduplicate before counting.

Screen Time: the summary a jury understands

Screen Time stores hourly usage in RMAdminStore-Local.sqlite and, when Share Across Devices is on, in RMAdminStore-Cloud.sqlite with data from other devices and family members. It is coarse, but it is Apple's own summary, and it is easy to present: "between 22:00 and 23:00 UTC, 47 minutes in TikTok on the device named 'Sam's iPhone'".

Always filter by ZCOREDEVICE. A parent's cloud store can contain a child's usage, and an iPad's usage can sit next to the iPhone's.

Network and interaction context

DataUsage.sqlite and netusage.sqlite show that an app or process used the network and when it was first and last seen. They are useful to establish that an app existed on the device during a period, even after deletion. interactionC.db adds who the owner interacted with through which app. Neither gives fine-grained usage, but both reach further back than most pattern-of-life streams.

What each acquisition level gives you

Pattern-of-life work is where the acquisition level matters most:

LevelWhat you get
Unencrypted or encrypted backupinteractionC.db and DataUsage.sqlite only; no knowledgeC, Biome, PowerLog or Screen Time
sysdiagnoseRecent PowerLog copies, unified logs for the last days, process lists
Full file system, BFUVery little: most pattern-of-life stores are protected by class keys unavailable before first unlock
Full file system, AFUEverything in this guide, including WAL files and Biome remote folders

If you only have a backup, say so plainly: the absence of pattern-of-life findings then reflects the acquisition, not the user's behaviour. A sysdiagnose taken at seizure is a valuable fallback when a full file system extraction is uncertain or delayed, because PowerLog copies and unified logs will cover the most recent days even if everything else ages out before an extraction becomes possible.

A workflow

  1. Inventory. List every source present, its date range and whether its WAL was collected. Write the inventory into your notes before analysis.
  2. Anchor the clock. Identify the device's time zone and changes (ZSECONDSFROMGMT in knowledgeC, Device.TimeZone in Biome, PowerLog time zone tables). Decide to work in UTC and convert only for presentation.
  3. Build device state first. Lock state, backlight and charging, from Biome and knowledgeC. This is the frame: nothing the user "did" can sit outside a lit, unlocked period without an explanation.
  4. Add app focus. From App.InFocus (and /app/inFocus), then check against PowerLog screen state and Screen Time hourly totals.
  5. Add context. Wi-Fi and Bluetooth connections (home network, car audio), media playback, Safari navigations, notifications.
  6. Add communications. Message sent and read times from sms.db and messaging apps, calls from CallHistory.
  7. Look for contradictions. Two independent sources disagreeing by minutes usually means a time zone or offset error; by seconds, it is normal logging latency.

Validation

Pattern-of-life data has no public specification, and parsers infer field meanings. Before relying on a stream in a report:

  • Test on a reference device of the same model family and iOS build: unlock, open a known app for a known time, lock, plug in, connect to a known Wi-Fi. Extract and check how each action appears, with what delay and in which stream.
  • Check the parser's notes. iLEAPP modules state what was validated and on which versions. APOLLO modules are tagged for older releases.
  • Verify a sample manually. Read a few records directly with sqlite3 or ccl-segb and compare with the parser output.

A single validated stream is worth more than ten unvalidated ones.

Interpreting common questions

"Was the phone in use at 21:40?" Look for an unlocked, lit interval covering 21:40 with an app in focus. Report the interval, the app, and which sources agree. If the screen was lit but the device locked, say so: that often means a notification woke the screen.

"Was the driver using the phone while driving?" Combine app focus and backlight with car connections (CarPlay, Bluetooth to the car), motion streams where present, and location data from the location guide. Note that CarPlay and Siri can produce activity without handling the phone.

"When did the person go to sleep?" The last unlock and the start of charging are indicators, not proof. Report them as such.

A worked example

The question is whether a driver was using his iPhone in the minutes before a collision recorded by a traffic camera at 17:42:10 UTC. The phone was seized at the scene, kept powered, and a full file system extraction was obtained in AFU state.

The inventory shows Biome streams covering the previous 18 days, knowledgeC.db sparse on this iOS 17 device, PowerLog with archives covering nine days, and Screen Time for four weeks. The device time zone was UTC+1 with no change that day.

Biome Device.Display.Backlight shows the screen lit from 17:39:52 to 17:42:31. _DKEvent.Keybag.IsLocked shows the device unlocked at 17:39:55. App.InFocus shows a messaging app in the foreground from 17:40:03 to 17:42:28, all local records, no remote ones. PowerLog screen state, after applying the time offset (+2.4 seconds), shows the same app frontmost at 17:40:05. Screen Time attributes 3 minutes of that app's use to the 17:00 hour on this device. sms.db holds no message in that window, but the messaging app's own database shows an outgoing message at 17:41:47. Biome shows a Bluetooth connection to the car's audio system from 17:21 onwards, and no CarPlay session.

The report states that the device recorded an unlock and continuous foreground use of the messaging app from 17:40:03 to 17:42:28 UTC, including an outgoing message at 17:41:47, while connected to the vehicle's Bluetooth, and that four independent sources agree within seconds. It does not state that the driver typed the message: that is for the other evidence.

Reporting

State the acquisition level and device state, list the sources used with their date ranges and tools, give times in UTC with the local offset, and describe the validation you performed. Avoid "the user did X". Prefer "the device recorded X", and add attribution evidence separately.

Common mistakes

  • Concluding "no app usage" from an empty knowledgeC /app/inFocus on iOS 16 or later.
  • Including Biome remote records in the iPhone's timeline.
  • Reporting raw PowerLog timestamps without the time offset.
  • Treating Screen Time cloud data as the examined device's activity.
  • Presenting app focus on a locked, dark screen as active use.
  • Forgetting WAL files, and then reporting that activity stopped days before seizure.

The timeline analysis guide explains how to merge these sources with messages, photos and locations using iLEAPP, APOLLO and a timeline viewer.

Related guides

05 · Location & Movement

iOS Location Forensics: Where Was the iPhone?

Every iPhone location source compared: Significant Locations, photo GPS, Wi-Fi networks, locationd caches, Biome and app data, with precision and retention.

Read guide