Home > Digital technologies > Cybersecurity > Forensic auditors: meet the Sherlock Holmes of cybercrime

Forensic auditors: meet the Sherlock Holmes of cybercrime

Published on 13 November 2026
Share this page :

Following a cyberattack, every clue counts. Deleted files, suspicious connections, traces left on servers… Forensic investigators play the part of Sherlock Holmes to piece together the facts, identify vulnerabilities and track down cybercriminals. Take a look behind the scenes of these digital investigations, where every byte can become a piece of evidence.

Image: Article: Forensic Auditors

Monday, 8.15 am

In a medium-sized company, none of the files will open. The ERP system has frozen, and the domain controller is no longer responding. On the accounts department’s screens, a message demands a ransom in cryptocurrency and threatens to publish stolen data. The IT department convenes the crisis team.

Management has four questions. How did they get in? How long have they been there? What did they take? Do they still have access? The answers will come from a digital forensics team tasked with reconstructing an attack in which the perpetrators have often covered their tracks.

The scenario is fictional. The methods described are those used in real investigations.

Securing the crime scene

An investigator’s worst enemy is not always the hacker. It is often the well-meaning administrator who restarts the server, runs an antivirus programme or «cleans» the infected machine. Each of these actions wipes out evidence.

Investigators follow a golden rule, the volatility order, formalised as early as 2002 in RFC 3227. First, the data that disappears most quickly is collected: CPU registers and cache, RAM, active network connections and running processes. Next come temporary files, followed by the hard drive, and finally backups and archives.

CERT-FR, the government’s incident response team, sets the same priorities: preserving logs and quickly identifying the extent of the impact.

A copy, never the original

Storage is never analysed directly. The investigator instead makes a bitmap image, often behind a hardware write blocker, using tools such as FTK Imager, dd or ewfacquire.

It immediately calculates a cryptographic hash, typically SHA-256. This hash ensures that the copy analysed six months later is identical to the one captured on the first day.

Every action is then recorded: who took what, when, where and why. This is the chain of custody. Without it, evidence that is technically flawless can be dismissed by a lawyer in court.

5 things to do in the event of an attack

  1. Alert the relevant teams and initiate the incident procedure
  2. Record each observation: time, message displayed, machine, action taken
  3. Preserving newspapers
  4. Don’t restart anything, don’t «clean» anything»
  5. Isolate the systems in consultation with the specialists

RAM: the most talkative witness

For a long time, the standard procedure was to switch off the computer to protect the hard drive. This is now considered a mistake. Many so-called malware programmes fileless They never settle on the hard drive: they reside in memory, either injected into a legitimate process or launched via an obfuscated PowerShell script. Cutting the power erases the culprit.

A RAM dump, analysed using a framework such as Volatility, reveals what the system is hiding: orphaned or hidden processes, injected DLLs, connections to a command-and-control server, and complete command lines. Encryption keys, or even plaintext credentials, are sometimes found there. A svchost.exe whose parent process isn’t services.exe? The investigator has just found their first suspect.

Deleted, but not gone

Deleting a file does not erase it. On NTFS, the system simply marks the entry in the Master File Table (MFT) as free. As long as the clusters are not overwritten, the data remains readable.

The tools for file carving Tools such as PhotoRec or the Autopsy modules go a step further: they scan unallocated space for file signatures – such as JPEG, PDF or ZIP headers – and reconstruct documents that the system has even forgotten the names of. Windows is, moreover, a formidable spy.

What Windows remembers

Windows also keeps a record of what has happened:

  • Prefetch : which executables ran, how many times and when
  • Amcache and Shimcache : traces of programmes, even those deleted since
  • LNK files and Jump Lists : documents opened by the user
  • USN Journal ($UsnJrnl) : every time a file is created, renamed or deleted
  • Snapshots (Volume Shadow Copies) : previous versions of the disk, which ransomware programmes are, incidentally, quick to delete.

One caveat regarding recent posts: on an SSD, the TRIM command instructs the controller to actually erase the freed blocks. This makes recovering deleted files significantly more hit-and-miss. Metadata, however, often survives.

Newspapers, or the art of chronology

An investigation is won on the timeline. Windows event logs form the backbone of the narrative, provided you know what to look for:

IDNewspaperWhat he says
4624 / 4625SecurityLogin successful / failed (type 10 indicates an RDP session)
4688SecurityCreate a process, including its command line if auditing is enabled
4720SecurityCreating a user account
7045SystemDeployment of a new service: a classic example of persistence
1102SecurityClearing the audit log

A burst of 4625 followed by a 4624 from the same IP address indicates a brute-force attack that ultimately succeeded. A 7045 at 3 am for a service with a random name often signals the arrival of a hijacked remote administration tool.

The investigator does not read these newspapers one by one. It consolidates all sources – logs, MFT, registers, browser history and firewall logs – into a super-timeline generated by tools such as Plaso (log2timeline). Hundreds of thousands of events are aligned to the second. Patient zero, the first suspicious connection and lateral movement can then be identified in just a few queries. However, the machines’ clocks must be synchronised: a poorly configured NTP server can throw the entire reconstruction off course.

On the network and in the cloud

Network logs do not tell the whole story. To track the attacker from one machine to another, the investigator turns to the network: PCAP captures where available, NetFlow data, and proxy, DNS and VPN logs. Tools such as Zeek or Wireshark can be used to identify a beaconing regular traffic to an unknown domain, or the exfiltration of several gigabytes on a Saturday evening.

With the move to the cloud, the crime scene has expanded. There are no longer any discs to seize from Microsoft 365 or AWS. You need to know how to analyse the unified audit log, CloudTrail or the Entra ID connection logs. And, above all, you need to be aware of their retention periods, which can sometimes be short depending on the licence held. An email transfer rule created quietly three months earlier may already have disappeared from the logs.

When the suspect stages the crime scene

Attackers are just as familiar with these techniques as investigators. The’anti-forensics is part of their toolkit: clearing logs using wevtutil cl, changing file dates (timestomping), tools loaded solely into memory, or the use of LOLBins, legitimate Windows binaries such as certutil or rundll32, which have been tampered with so as not to trigger the antivirus.

But erasing one trace often leaves another. Clearing the security log actually triggers event 1102, and a six-hour gap in the logs is in itself a clue. Timestomping is revealed by comparing the two sets of timestamps in the MFT: $STANDARD_INFORMATION, which is easy to modify, and $FILE_NAME, which is much less accessible. When the former claims a file dates from 2019 and the latter says yesterday, the investigator knows they are being lied to. Just like in Conan Doyle’s stories, it is sometimes the dog that didn’t bark that solves the case.

From investigation to response

In the corporate world, forensic analysis and incident response go hand in hand. Indeed, in the English-speaking world, the two have been combined under a single acronym: DFIR, for Digital Forensics and Incident Response. The challenge is not just to understand, but to act swiftly: to contain the threat, eradicate the persistence and restore the system, without destroying the evidence in the process.

To ensure everyone is on the same page, the teams map the techniques observed against the reference framework MITRE ATT&CK : initial access, execution, persistence, privilege escalation, lateral movement, exfiltration. The extracted indicators of compromise – hashes, IP addresses and domain names – are fed into the SIEM and EDR systems to verify that no other machines are affected. Tools such as Velociraptor or KAPE enable these artefacts to be collected from hundreds of workstations at once.

The investigation concludes with a report. This is provided to management, the cyber insurer, the CNIL in the event of a personal data breach that must be reported within 72 hours, and sometimes to the court. It must be factual, verifiable and understandable to a non-technical reader. It is not the most spectacular part of the job, but it is often the most decisive.

A profession that requires patience and rigour

Forget those TV scenes where a hooded hacker is unmasked in three clicks. A forensic auditor spends hours sorting through, cross-referencing and documenting. They need a solid understanding of file systems – both Windows and Linux – as well as network protocols and cloud architectures. They must also have a grasp of scripting – whether Python or PowerShell – to automate whatever can be automated. And above all, they must keep a cool head when the whole company is waiting for an answer.

Good news for system administrators, network engineers and SOC analysts: they already have much of the necessary knowledge. What they lack is the methodology: knowing how to collect evidence without compromising it, knowing where to look, and knowing how to prove it.

Training

This is precisely the focus of several ORSYS training courses. Forensic analysis (3 days) provides an introduction to digital forensics techniques: evidence collection, analysis of traces on a compromised system, and reconstruction of the attack scenario. Forensic Analysis and Security Incident Response (4 days) Goes further: in-depth analysis of breaches, identification of attackers’ techniques and mastery of the stages of the incident response process to contain the threat and secure the IT system.

Enough to enable him, when the time comes, to go from being a helpless witness to an investigator.

Our experts

Made up of journalists specialising in IT, management and personal development, the ORSYS Le mag editorial team [...]

field of training

associated training