← all systems · see it in The System
Digital forensics
A branch of forensic science that recovers, examines and analyses material found in digital devices (src:wp:digital-forensics): the practice of preserving and copying what devices, networks and providers hold, examining and interpreting it against competing hypotheses, and reporting it for a court, an investigation or an organisation (src:nist-sp800-86; src:nistir-8354). Walked as the practice and the people, tools, standards and labs that do it; at least 11,000 US organisations by NIST's lower-bound estimate.
Properties published scienceour proposal
Sourced fields, and our own readings.
| aliases | Digital forensics and evidence handling |
| born at | 1984 (FBI Computer Analysis and Response Team), contested |
| born precision | year |
| boundary | Inside: the forensic process and its practitioners (first responders who seize and preserve, examiners and analysts, reviewers), their tools (write blockers, imagers, parsers, carvers, hash checks), their procedures, notes, chain-of-custody records and reports, and the labs, standards bodies and test programmes that hold them to account (SWGDE, NIST CFTT, ISO/IEC 17025 accreditation). Outside: the devices and the people whose trail is examined; the person who altered it (anti-forensics); the police investigation, prosecutors and courts that ask the questions and decide (the edge between investigator and examiner is drawn differently by different agencies; digital media investigators sit on it); incident response, which forensic work is part of inside organisations; providers that hold third-party data. |
| nested level | A technology-era practice (rung 12), a branch of forensic science; carried by people (Human) and run on computers; part of incident response inside organisations and part of criminal investigation in policing. |
| scale | social |
Measures (24) published science
| US digital forensic organisations, lower bound (capture-recapture) | 11,000 (95% CI 9,900-12,600) organisations |
| publicly funded US crime labs (BJS) | 409 labs |
| unique US digital evidence groups on five lists | 5,457 groups |
| digital forensic tools tested by NIST | over 250 tools |
| law-enforcement digital forensic backlog | 6-12 months commonplace; often 2 years; up to 4 years time |
| mobile phone backlog, England and Wales | 9-12 months in some places; 3 months nationally |
| onset of SSD wear levelling and garbage collection after deletion | within the first minute; 0 of 70 files recovered with original hash minutes; files |
| growth in demand for digital forensic analysis (England and Wales) | 11-16% over three years percent |
| crimes with a digital component (England and Wales strategy) | over 90 percent |
| processes requiring accreditation that achieved it (England and Wales) | less than 20 percent |
| inter-examiner consistency, same file and context (Sunde and Dror 2021, 53 examiners) | 0.667 (Krippendorff's alpha, low/inadequate) |
| FBI RCFL data processed, 2013 | 5,973 TB from 7,273 examinations (up 40% on 2011) |
| pages equivalent from one phone extraction | up to 33 million paper pages |
| devices in a normal inquiry | up to five devices |
| written alibi corroboration after incriminating evidence vs none | 47% vs 81% percent |
| OCT volumetric scan of a micro-SD card | ~37 s |
| OCT resolution (lateral / axial) | ~7 / ~4 micrometres |
| NTFS file access time resolution (FAT: 1 day) | 100 ns |
| head start of conventional over digital forensics | around 90 years |
| cases with encrypted devices that go unprocessed (estimate) | about 60 percent |
| prosecutors' use of digital forensic and third-party data across six crime types | 60-80% of the time percent |
| UKAS surveillance visits to accredited DF units, Sep 2018-Apr 2019 | 29 visits |
| anti-forensics publications analysed | 123 papers |
| likelihood ratio, Hong Kong internet auction fraud cases (Bayesian networks) | 164,000 in favour of prosecution LR |
Open contentions (4) open contention
Where sources or readings disagree. Both stay on the map.
- C-1600 When did digital forensics begin?
booked: The map dates the practice to the 1980s police units, and keeps the case that Stoll's 1986 hunt and the 1988 Morris worm mark the start.
the other reading: Stoll's 1986 pursuit of Markus Hess and Spafford's work on the 1988 Morris worm mark the beginning of systematic digital forensic investigation (Overill and Collie 2021); a pre-history before infancy, childhood and adolescence, from a participant's personal history (Pollitt 2010); the first computer crime law in 1978 (Florida) marks the start of the legal need - C-1601 Can a digital forensic tool be given an error rate?
booked: The map books reliability through testing and keeps both views: single error rates mislead for tools that fail systematically, and courts ask for them anyway.
the other reading: error rates are inappropriate or misleading for digital processes, which err systematically (NIST IR 8354); Daubert asks for known error rates, and they can be measured by repeated controlled tests (Ismail et al. 2025) - C-1602 Is the general lab standard (ISO 17025) right for digital forensics?
booked: The map records the standard as in force in England and Wales and keeps both the case for it and the case that digital work needs its own.
the other reading: ISO 17025 is not fit for purpose for digital forensics; a specific standard is needed (as reported in Reedy 2023); ISO 17025 accreditation is critical to competent, decentralised digital forensic science (as reported in Reedy 2023) - C-1603 Should digital forensic findings come with a probability?
booked: The map records that findings are mostly reported without a number, and keeps open whether they should carry one.
the other reading: attempts to quantify uncertainty should be abandoned in digital forensics (Horsman 2020, as quoted by Biedermann and Kotsoglou); probability is a general measure of uncertainty and applies to digital evidence as to any other; exceptionalism is unwarranted (Biedermann and Kotsoglou 2020); Bayesian networks, likelihood ratios and complexity models already quantify digital findings in real cases (Overill and Collie 2021)
Partners and parts published science
Built from / builds (9)
- Incident response
- Cryptographic hash function
- Documentation and external affirmation (third type: certificates, chain of custody)
- Digital forensic examiners and first responders
- Write blockers
- Forensic software tools (imagers, parsers, carvers, timeline and link analysis)
- Digital forensic laboratories (crime labs, RCFLs, ICAC labs, corporate and private labs)
- Anti-forensics against digital forensics
- Image forgers and detectors
Sits inside / holds (4)
Trades with (valence) (14)
- Photography via sys:sensory-environment · Digital forensics gain
- Insider threat via sys:computer-network · Digital forensics gain
- Cybercrime and its underground economy via sys:internet · Digital forensics gain
- Human (Homo sapiens) via sys:human-society · Digital forensics gain
- Law (rules and law) via sys:human-society · Digital forensics mixed
- Operating system via sys:computer-hardware · Digital forensics gain
- Police via sys:human-society · Digital forensics gain
- Cryptographic hash function via sys:written-record · Digital forensics gain
- Incident response via sys:written-record · Digital forensics gain
- Threat actor via sys:computer-hardware · Digital forensics loss
- Computer hardware via sys:computer-hardware · Digital forensics mixed
- Cloud computing via sys:law · Digital forensics gain
- Intruder (burglar and trespasser) via sys:built-environment · Digital forensics gain
- Forensic science via sys:written-record · Digital forensics gain
expressed by (23)
- Generation/Synthesis
- Templating / Standardization
- Quality Improvement/Refinement
- Interoperability/Transduction (Bridging)
- Distributed Autonomy/Decentralized Execution (Subsidiarity)
- Sensing/Scanning
- Search/Foraging
- Mapping/Modeling
- Uncertainty Reduction/Learning
- Active Experimentation / Probing (Optimal Experimental Design)
- Multi-Scale Data Fusion / Triangulation
- Trail-Marking / Externalized Memory (Stigmergy)
- Frontier Outposts / Sensing Extension
- Hypothesis Portfolio / Multiple Hypothesis Tracking
- Social Learning / Public Information Use
- Reactive Response
- Proactive Defense
- Physical/Virtual Barrier
- Validation of Identity/Intention
- Validation of Effectiveness
- Redundancy & Diversity
- Triage & Prioritized Allocation
- Quality Control & Error Correction
floor filled by (16)
- Incident response
- Human (Homo sapiens)
- Law (rules and law)
- Operating system
- Police
- Cryptographic hash function
- Computer hardware
- Cloud computing
- Documentation and external affirmation (third type: certificates, chain of custody)
- Digital forensic examiners and first responders
- Write blockers
- Forensic software tools (imagers, parsers, carvers, timeline and link analysis)
- Digital forensic laboratories (crime labs, RCFLs, ICAC labs, corporate and private labs)
- Forensic science
- NIST CFTT
- SWGDE
Kinds (8)
- Computer forensics
- Mobile device forensics
- Network forensics
- Memory forensics
- Forensic data analysis
- Electronic discovery (eDiscovery)
- Remote and cloud acquisition
- Anti-forensics (counter-forensics)
Notes published science
- A solid-state drive can destroy the evidence on it by itself: after files were deleted, the drive's own housekeeping started within the first minute, and none of 70 files came back with its original fingerprint (hash), even though the investigators changed nothing (src:hadi-2023-ssd-trim). NIST warns that two images of the same SSD taken a little apart may not match for the same reason (src:nistir-8354).
- Showing people digital evidence can rewrite their memory-based testimony: alibi witnesses who were told of incriminating camera footage backed the alibi in writing only 47% of the time, against 81% with no evidence (src:sauerland-2023-alibi). The record and the memory are not independent once they meet.
- Nobody knew how many digital forensic labs the US has, so NIST counted them the way ecologists count fish: capture-recapture across five overlapping lists gave at least 11,000, against 409 public crime labs (src:nistir-8354).
Layers
The ways it takes in its world, where they meet, and how it acts back, each from L1 (where it meets the world) up.
digital forensics as forensic examiners describe it: the trail left on digital devices up to the decision it serves (reader's view)
- L1: The digital trail: devices, storage media, memory, networks and provider records, where every interaction leaves artifacts
- L2: Identification, seizure and protection: recognise the sources, write-block, shield, decide on volatile data
- L3: Acquisition and integrity: forensic image or logical copy, verified by hash; chain of custody begins
- L4: Examination: recover deleted data, parse, identify and extract artifacts
- L5: Analysis: reconstruct events, build timelines, weigh alternative hypotheses, check for anti-forensics and bias
- L6: Reporting, review and standards: the report, notes, chain of custody, validation, accreditation
- L7: The decision served: a court's verdict, an investigation's next step, an organisation's response
the boardroom (shared floors)
- L4: Examination: recovery of deleted data, parsing the file systems, identifying and extracting artifacts from the acquired data while protecting its integrity
- L5: Analysis and interpretation: timelines, link analysis, competing hypotheses, attribution, checks for anti-forensics and for bias
- L6: Reporting, review and quality: the report in lay terms, contemporaneous notes, chain of custody, peer review, tool validation, accreditation (ISO/IEC 17025), lab management
- L7: The decision served: the court (admissibility, the trier of fact), the investigation (leads), the organisation (incident response, discipline, compliance)
way in: Anti-forensics: what the subject did to the evidence
- L1: Actions on the device: destroying, hiding, eliminating sources of, or counterfeiting evidence
- L2: Traces altered: wiping, timestamps changed, the clock set wrong, logs edited, steganography
- L3: Planted content that tools, examiner and an LLM read differently (the perception gap)
- L4: Caught or missed in examination: detection rests on the examiner's skill
way in: Artifact knowledge: what each trace means in each software version
- L1: Operating systems and applications whose behaviour writes artifacts, changing with every version
- L2: Practitioners try out methods and test artifact behaviour
- L3: Shared through blogs, white papers, videos and catalogues (Artifact Genome Project, curated crowdsourced catalogues)
- L4: Built into tools' parsers and the examiner's reference
way in: Case context from investigators (task-relevant and task-irrelevant)
- L1: The investigators' case knowledge
- L2: The submission form's 'case information' field, with no guidance on what to leave out
- L4: Exposure sequenced by biasing power, objectivity and relevance (LSU-E, a general forensic method)
way in: Chip-level storage (monolithic storage devices, damaged media)
- L1: Interface, controller and flash on one board under an opaque layer
- L2: Non-destructive inspection by near-infrared OCT to find the pins
- L3: Targeted laser ablation exposes only the pins needed, then the flash is read
way in: Cloud and third-party provider records
- L1: Data held by companies, not on the seized device
- L2: Legal process: warrants and subpoenas to the provider; remote acquisition cannot be write-blocked
- L3: Returns received, often slow
way in: Demand: cases, devices per case, data per device
- L1: Crimes with a digital component: over ninety percent (England and Wales strategy)
- L2: More devices per case and more data per device
- L3: The queue: backlogs of six months to a year, up to four years
- L4: Risk-based triage and service levels
way in: Human accounts: suspects' statements, alibis, witnesses' and victims' accounts (Human: memory and testimony)
- L1: Memory of the event in the people involved, shaped by expectations and by what they learn later
- L2: Interviews: victim interviews feed offender interviews and evidence plans
- L3: Statements and alibis on record, to be cross-checked
way in: Law and authorisation (warrants, rules of evidence, privacy law)
- L1: Statutes and rules: Federal Rules of Evidence, PACE, ECPA, Computer Misuse Act
- L2: Authorisation and a warrant before the search; all contents preserved so only authorised portions are examined
- L4: Seizure limits: US federal law restricts seizure to items of obvious evidential value
- L5: Expert opinion must rest on sufficient facts and reliable methods reliably applied (FRE 702)
way in: Mobile devices (phones, tablets, GPS units)
- L1: Phones that keep talking to the network: call data, messages, location from GPS and cell-site logs, proprietary storage
- L2: Isolation: a Faraday shield or bag at seizure stops further radio traffic
- L3: Extraction: usually without write blocking, often by loading a tool onto the phone
way in: Network traffic and logs
- L1: Traffic and logs on the network, local and wide-area
- L2: Capture points such as IDS sensors see only part of a connection
- L3: Correlation by time: synchronised clocks, recorded drift
way in: Non-volatile storage (hard drives, SSDs, flash drives, memory cards)
- L1: The stored trail: files, metadata, deleted content left in unallocated space, and on SSDs blocks the drive may erase on its own
- L2: Protection: a hardware write blocker between drive and examiner's computer suppresses commands that would change the drive
- L3: Acquisition and integrity: a sector-level forensic image, hashed with the original and compared
way in: The examiner's own expectations and judgement (inner)
- L1: The examiner as a person: knowledge from training, vendors, blogs, books; expectations formed from context
- L2: Quick judgement under time pressure ('golden hour' decisions, personal judgment)
- L4: What is observed is shaped by context: more traces found under a guilt context
way in: Tool testing results (CFTT reports, CFReDS reference images)
- L1: Reference images built so anomalies show (every test sector stamped with its address)
- L2: Test cases run against each tool function
- L3: Reports published (via NIJ and DHS); over 250 tools tested
way in: Volatile memory and live system state (registers, cache, RAM, process table, connections)
- L1: Volatile data, lost at power-off or with time, ranked by an order of volatility: registers and cache first, archival media last
- L2: Decide fast whether to preserve the volatile data, by criteria set in advance
- L3: Capture with trusted tools run from protected media, recording clock drift and who was present
way out: Containment advice in incident response
- L1: Cables disconnected, power removed, host shut down
- L7: Containment decision with the incident response team, weighing risk against the integrity of evidence
way out: Evidence put to suspects and used with victims in interviews
- L1: Spoken and shown to the person interviewed
- L5: Strategic disclosure planned: usually late in the interview, to challenge the account
way out: Expert testimony in court
- L1: Spoken to judge and jury, who may over-trust or undervalue it
- L7: Testify as an expert, if the method passes the judge's gate
way out: Improvements: procedures, and logging set to record more
- L1: Auditing, logging and intrusion detection altered to record more for future investigations
- L6: Formal reviews after each major event
way out: Knowledge shared back to the community
- L1: Curated catalogues of artifacts
- L6: Feedback on usefulness and validity circulated
way out: Leads and new collection
- L1: A fresh search warrant on the new information
- L5: Actionable information identified
way out: The written report, in lay terms
- L1: The report in the case file, read by investigators, lawyers, managers
- L6: Report written for its audience