Abstract

Digital evidence is valuable only when its integrity can be demonstrated and its meaning defended. This article examines forensically sound preservation, supported by write-blocked imaging, cryptographic hashing and a documented chain of custody, together with the interpretation of three Windows artefacts: Prefetch, NTUSER.DAT and the System Resource Usage Monitor (SRUM). A correlation workflow is proposed that keeps factual observation separate from professional inference.

1.  Introduction

Digital evidence is now central to investigations ranging from financial fraud and identity theft to intrusion and homicide, because almost every human activity leaves a trace in stored or transmitted data [1]. Yet its persuasive power is fragile. Electronic records are abstract, frequently circumstantial and easily altered, sometimes by the act of examining them. A case is therefore more likely to fail because the handling of evidence cannot be trusted than because the evidence itself is wrong.

This article first describes how evidence is preserved so that its authenticity withstands judicial scrutiny. It then explains how Prefetch, NTUSER.DAT and SRUM help establish which programs were executed, under which account, when, and with what resource footprint, before proposing a correlation workflow and discussing its limitations.

2.  Preservation and Integrity of Digital Evidence

2.1  Order of Volatility and Forensic Imaging

Evidence is collected in decreasing order of volatility. Memory, running processes and active network connections disappear when power is lost and are therefore captured before disk contents and archived data [2]. Before anything is moved, the scene, screen state and cable connections are photographed, and mobile devices are isolated from wireless networks, typically in a Faraday enclosure, to prevent remote wiping.

Examination is never performed on the original. A hardware or software write blocker prevents data from being written back to the source medium while a bit-for-bit forensic image is created with a tool such as FTK Imager. The examiner then works exclusively on the image, and the sealed original remains as seized. Forensic soundness does not demand that nothing ever changes, since live memory capture alters a running system; it demands that every change is minimal and documented.

2.2  Hash Verification

A cryptographic hash function converts data of any length into a fixed-length digest, and altering even a single bit produces an entirely different value. Hashing the source at acquisition,hashing the resulting image, and recomputing the digest at each later stage supplies objective proof that content is unmodified. Because practical collision attacks exist against MD5 and SHA-1, which many acquisition tools still report, a SHA-256 value should also be recorded. The governing rule is simple: the acquisition hash and the verification hash must match.

2.3  Chain of Custody and Admissibility

Hash values show that data are unchanged; the chain of custody shows who held the data, when, and why. Every transfer is recorded with the case and item number, date and time, releasing and receiving parties, purpose, storage location, item condition and both signatures. Typical failures include missing timestamps, undocumented handovers and absent receiver signatures. In India, electronic records have been admitted through Section 65B of the Indian Evidence Act, 1872, now carried forward as Section 63 of the Bharatiya Sakshya Adhiniyam, 2023 [4], and the Supreme Court in Anvar P.V. v. P.K. Basheer treated the accompanying certificate as a condition of admissibility [3]. A broken or poorly documented chain therefore exposes even technically flawless findings to challenge.

3.  Windows Artefacts as Behavioural Evidence

Artefacts are traces of user or system activity generated automatically by Windows and its applications, and they frequently persist after the user deletes the files that produced them. No single artefact tells the whole story; each must be combined with others.

3.1  Prefetch

Prefetch is a performance mechanism that records the files an application loads so that later launches are faster. The resulting .pf files reside in C:\Windows\Prefetch and are named after the executable and a hash of its path, as in CHROME.EXE-3F4A6A9C.pf. Their content includes the executable name, run count, last execution time, referenced DLLs and volume information. An examiner can therefore establish that a program ran, how often, and when it last ran, even after the executable has been removed. An entry such as MIMIKATZ.EXE strongly indicates that a credential-dumping tool was executed. Prefetch nevertheless proves execution only; it neither establishes intent nor identifies the responsible account, and the feature may be disabled on some Windows editions.

3.2  NTUSER.DAT

NTUSER.DAT is the registry hive that stores configuration and activity for one user account and is located in that user's profile folder, C:\Users\<Username>. Because every account has its own hive, it attributes activity to a person rather than to the machine, which is decisive on shared computers. Its principal evidentiary keys are RunMRU (commands typed into the Run dialogue), RecentDocs (recently opened files), TypedURLs (addresses typed into the legacy Internet Explorer address bar) and UserAssist (programs launched through the graphical shell, with run counts and timestamps, stored under ROT13-encoded names). While the account is logged in the hive is locked, so it is best extracted from the forensic image together with its transaction logs (.LOG1 and .LOG2), since a hive that was not cleanly closed may hold uncommitted entries.

3.3  System Resource Usage Monitor (SRUM)

The System Resource Usage Monitor, introduced in Windows 8, records how applications consume system resources. Its data are held in an Extensible Storage Engine database, C:\Windows\System32\sru\SRUDB.dat, and include application identity, user security identifier (SID), network bytes sent and received, CPU and energy usage, and timestamps. Whereas Prefetch answers whether a program ran, SRUM describes what it did while running. An unusually large data transfer, for instance, may indicate exfiltration, and the database can retain evidence of applications that no longer exist on disk. Retention is limited to weeks or a few months, so prompt acquisition matters.

4.  Analytical Workflow and Correlation

The procedure follows the order in which the artefacts naturally build a narrative. First, Prefetch is parsed with a tool such as PECmd to identify executed programs and timestamps. Second, NTUSER.DAT is processed with a registry parser such as RegRipper or the Python library regipy to link execution and file-access traces to a specific account. Third, SRUM is parsed, for example with SrumECmd, ideally alongside the SOFTWARE hive so that identifiers can be resolved, to add network and resource dimensions. Finally, all timestamps are normalised to UTC and merged into one timeline. Event log records such as 4624 (successful logon), 4688 (process creation) and 6008 (unexpected shutdown) provide an independent anchor, for example by confirming which account was active when a program ran.

Three safeguards apply throughout. Parsing is performed on exported copies or a mounted image, never on the live original, and each exported artefact is hashed and logged. Tools and versions are recorded so that results can be reproduced. Lastly, observation and interpretation are kept apart: the examination log records only what was seen, such as SRUM records for an unfamiliar executable on a given date, whereas the significance of that observation is stated separately, with reasoning, in the analysis section of the report. This separation preserves the objectivity of expert testimony.

5.  Limitations and Challenges

Artefact analysis carries inherent uncertainty. Timestamps depend on system clocks that may be inaccurate or deliberately altered. Anti-forensic activity, including deletion of Prefetch files, clearing of event logs and timestamp manipulation, can remove or distort evidence, although redundancy among artefacts often exposes such tampering. Finally, evidence of execution is circumstantial: it places a program on a system under an account but does not by itself prove who was at the keyboard. Conclusions must therefore be expressed with proportionate caution and corroborated independently.

6.  Conclusion

Reliable digital forensics rests on two inseparable foundations: a preservation process that can be verified and an interpretation that can be reproduced. Write-blocked imaging, hash verification and a meticulous chain of custody establish that evidence is what it purports to be, while correlated analysis of Prefetch, NTUSER.DAT and SRUM converts it into a coherent account of user and system behaviour. Future work should automate artefact correlation and hash-verified reporting so that this rigour scales with data volumes.

References

[1]  Casey, E. (2011). Digital Evidence and Computer Crime (3rd ed.). Academic Press.

[2]   Kent, K., Chevalier, S., Grance, T., & Dang, H. (2006). Guide to Integrating Forensic Techniques into Incident Response (NIST SP 800-86). NIST.

[3]  Anvar P.V. v. P.K. Basheer, (2014) 10 SCC 473 (Supreme Court of India).

[4]   Government of India. (2023). The Bharatiya Sakshya Adhiniyam, 2023 (Act No. 47 of 2023).