What Is Fileless Malware? How It Works, Detection & Best Defense Solutions

Most people picture malware as a file: something you download, scan and delete. Fileless malware breaks that picture. It runs in memory and through tools that Windows already trusts, so the usual warning signs can be faint. Yet it still leaves clues in process trees, logs, the registry and network traffic. CrowdStrike's 2026 Global Threat Report found that 82% of detections in 2025 were malware-free, which shows how much attackers now lean on trusted tools and stolen access. This guide explains what fileless malware is, how it works, how to spot it, and which defenses deserve your budget.
TL;DR
Fileless malware is a loose label for attacks that run in memory or through trusted system tools, often with little or no new executable on disk.
"Fileless" does not mean invisible, and it does not always vanish after a reboot. Registry entries, WMI subscriptions and scheduled tasks can keep an attack alive.
Fileless, living-off-the-land (LOTL) and malware-free are overlapping ideas, not synonyms.
Strong detection combines command-line and script logging, Sysmon or EDR process telemetry, identity logs and network data.
Prevention is layered: patching, phishing-resistant MFA, least privilege, application control, script restrictions and a tested response plan.
Penetration testing can reduce certain entry-point weaknesses. It does not detect or stop fileless malware by itself.
What is fileless malware?
Fileless malware is an attack technique that runs malicious code in memory or through trusted built-in tools, such as PowerShell or WMI, instead of relying on a traditional executable file. Some fileless threats still touch disk, registry or logs, so detection depends on watching behavior, not just scanning files.
![]() Find the Weak Doors Before Attackers Do Many fileless attacks begin with an ordinary break-in: a stolen login, a flawed web app or an exposed API. Articsledge offers scoped application, API and SaaS security testing that helps you find and prioritize those entry points first.
Request a Penetration Testing Quote Every engagement is scoped individually and depends on specialist availability and agreed terms. |
Table of Contents
What Is Fileless Malware?
Fileless malware is a threat that does not depend on a malicious file sitting on disk. Microsoft explains that there is no single agreed definition, and that the term is sometimes stretched to cover families that do use files. The core idea is simple: the attacker avoids dropping a classic executable and instead runs code in memory or through tools already on the machine.
That matters because many older defenses were built around files. A scanner can inspect an executable before it runs. It has a harder job when the "payload" is a script block, a registry value or code injected into a healthy process. The threat is still real software activity, though, and it leaves traces. Microsoft has said plainly that being fileless does not mean being invisible.
Why the label is slippery
Microsoft sorts these threats by how much file activity they involve. Type I has no file activity at all, such as code delivered straight into kernel memory or malicious firmware. Type II has indirect file activity, such as a PowerShell command stored in the WMI repository. Type III needs files to operate, even though much of the behavior is fileless. Type I is described as uncommon, so most real-world cases sit in the second and third groups.
Related terms add to the confusion. A malware family may use fileless steps in one stage and ordinary files in another. A rootkit hides itself deep in the system, but it is not automatically fileless. A threat can be both, neither or only partly one of these.
Myths versus facts
Myth: Fileless malware never touches the disk. Fact: Many chains write files at some stage, and registry or WMI storage still lives on disk.
Myth: It always disappears after a reboot. Fact: Persistence through the registry, WMI or scheduled tasks can survive restarts.
Myth: Antivirus cannot see it. Fact: Modern products use AMSI, behavior monitoring and memory scanning, though no control catches everything.
Myth: Every PowerShell command is an attack. Fact: Administrators use PowerShell and WMI daily. Context decides whether an event is malicious.
Who is most exposed
Windows-heavy organizations face the most documented cases, because PowerShell, WMI and many LOLBins ship with the operating system. Teams with weak logging, broad admin rights and slow patching are exposed most, whatever their size. Small firms are not safe simply because they are small. They often lack the telemetry needed to notice quiet activity.
How Does Fileless Malware Work?
There is no single script for a fileless attack. The sequence below is an illustration, and real incidents skip, reorder or repeat steps.
Initial access: a phishing email, a stolen password, an exposed service or a vulnerable application gives the attacker a foothold.
Execution: a document macro, shortcut file or script launches a built-in interpreter such as PowerShell, mshta or a script host.
Native-tool abuse: trusted utilities download, decode or launch the next stage, so each step looks like normal Windows behavior.
Memory payload: the main code is loaded into memory, sometimes inside a legitimate process.
Persistence (optional): the attacker may store a launcher in the registry, a WMI event subscription or a scheduled task so it can restart.
Identity abuse: stolen tokens, passwords or admin rights extend access without any new malware at all.
Lateral movement and command and control: remote administration tools and outbound connections let the attacker move and receive instructions.
Impact: data theft, fraud, surveillance or ransomware deployment.
Notice how many steps could be performed by a real administrator. That overlap is the whole point. Defenders therefore watch the chain, not a single event. A harmless-looking script launched by an Office document, followed by an odd network connection, tells a very different story than the same script run by a systems engineer.
Speed matters too. CrowdStrike reported an average eCrime breakout time of 29 minutes in 2025, the time from first access to moving sideways. Fast movement shrinks the window for manual review, which is why automated alerting and a rehearsed response matter. Wider fallout can resemble a cyber pandemic, where one weakness spreads disruption across many connected systems.
Fileless Malware vs. Traditional Malware vs. Living-off-the-Land Techniques
These three ideas overlap but are not the same. Traditional malware depends on a malicious file. Fileless malware avoids a classic payload file. Living off the land (LOTL) means abusing legitimate tools and features, whether or not malware is involved. A pure LOTL intruder may use only stolen credentials and built-in admin tools, so it is not "malware" in the usual sense.
Aspect | Traditional malware | Fileless malware | Living-off-the-land activity |
Core idea | Runs a malicious executable or library | Runs code in memory or via script, registry or WMI | Uses trusted tools and accounts for malicious aims |
Disk artifacts | Usually clear files and hashes | Fewer files; registry, WMI or logs may hold traces | Tool use, logs and command lines |
Persistence | Services, run keys, dropped files | Registry, WMI subscriptions, scheduled tasks, or none | Valid accounts, remote tools, scheduled tasks |
Examples | Classic trojans and droppers | Poweliks, Kovter, PoshSpy | Astaroth's tool-only chain; admin-tool abuse |
Detection signals | File scanning, hashes, sandboxing | Behavior, memory scanning, script content | Baselines, process ancestry, identity anomalies |
Key limit | Signatures miss new variants | Memory-only evidence can vanish on reboot | Hard to separate from real admin work |
The takeaway: LOTL is a method, fileless is a style of execution, and malware-free is a statistic about detections. CrowdStrike's 82% figure counts detections that did not involve malware. It is not a measure of fileless malware specifically. Government guidance on identifying and mitigating living-off-the-land techniques is a useful companion for defenders.
Types of Fileless Malware and Common Techniques
Microsoft groups fileless threats by the place they run: exploits, hardware, and execution or injection. The techniques below are the ones defenders meet most. Each is mapped to MITRE ATT&CK where a verified entry exists.
Script-based execution (PowerShell and script hosts)
PowerShell is a command and scripting interpreter, tracked as T1059.001 under the Execution tactic. JavaScript, VBScript and HTA files run through script hosts such as wscript and mshta follow the same pattern. Scripts can run from memory, from a registry value or from a scheduled task. Not every script is hostile, so logging the content matters.
WMI-based execution and persistence
Windows Management Instrumentation (WMI) lets administrators manage systems locally and remotely. Attackers can use it for execution, tracked as T1047. They can also register WMI event subscriptions to store and trigger code. MITRE's detection guidance for WMI abuse looks for a chain of process creation, command execution and, for remote use, a matching network connection.
Registry-assisted payloads
The registry can store encoded payloads or launch commands. Modification of it is tracked as T1112, which MITRE lists under both Persistence and Defense Impairment. Registry use is classified by Microsoft as a storage location inside script-based threats, not as a separate category.
Process injection
Process injection runs code inside another process, which can hide it behind a trusted name. MITRE tracks it as T1055, with 12 sub-techniques such as DLL injection and process hollowing. It is a memory technique that may or may not accompany files.
Reflective code loading
Reflective code loading, T1620, loads code into a process's own memory without the normal operating system loader. MITRE notes that it cannot be easily mitigated with preventive controls because it abuses normal system features. It points toward detection of unusual memory behavior, such as memory made executable inside a process that does not normally do that.
LOLBins and macro-based chains
Living-off-the-land binaries (LOLBins) are legitimate programs that attackers repurpose, such as regsvr32, certutil, BITSAdmin and mshta. Office macros can start the chain: Microsoft lists macro-based threats, including the qkG ransomware, as a fileless-adjacent category. Some tools are inherently memory-focused, while others only overlap with fileless patterns. Treat the labels as descriptions of behavior, not as a verdict. Tools that guide AI helpers with command access create a similar risk, which is why agentic AI security now belongs in the same conversation.
Common Entry Points and Attack Vectors
A fileless attack still has to get in. Match each entry route to a control that closes it.
Entry point | How it helps the attacker | Controls that help |
Phishing and malicious documents | Persuades a user to open a shortcut, script or macro | Mail filtering, macro blocking, awareness training |
Credential compromise | Gives valid access without malware | Phishing-resistant MFA, conditional access, monitoring of sign-ins |
Vulnerable services and applications | Allows code execution in memory | Patching, vulnerability management, exposure reduction |
Browser-related paths | Delivers scripts or exploit content | Browser hardening, updates, web filtering |
Remote administration misuse | Lets attackers move with approved tools | Least privilege, just-in-time access, remote-tool allow lists |
Web and API weaknesses | Provides a foothold on servers | Secure development, authorization testing, scoped penetration tests |
Trusted software and update paths | Abuses established trust | Vendor review, code signing checks, change monitoring |
Human tricks remain the most common door, so understanding social engineering helps non-technical staff as much as engineers. Attackers also target the software around the endpoint. Poorly designed APIs and web apps can give them a server-side foothold. Newer AI features create fresh doors too: an assistant that reads untrusted text can be steered by prompt injection into running commands on a user's behalf, and the general risks are covered in this LLM security guide.
![]() Test the Doors Attackers Actually Use Weak authentication, broken authorization and exposed APIs can hand attackers a first foothold. Scoped web application and API penetration testing checks how those controls behave under authorized attack and reports the evidence your developers need.
Discuss an Application or API Assessment Scope, rules of engagement and timing are agreed before testing begins. |
Real-World Examples of Fileless Malware
These cases are well documented. None is a perfect "pure" example, which is the lesson in itself.
Poweliks (2014)
Researchers at G Data described Poweliks in July 2014 as persistent malware without a file. It stored encoded code in the registry and used an autostart entry with a hidden, non-ASCII name so common tools could not open it. It reportedly arrived through a malicious Word document, so an initial file existed. Detection signals include odd registry autostart values and a script interpreter launching from them. The lesson: "fileless" often describes the persistence, not the whole chain.
Kovter (observed from 2016)
Kovter was click-fraud malware. Microsoft's analysis found it registering a random file extension with a shell "open" command in the registry, so opening a file with that extension ran mshta to execute JavaScript, which loaded the main payload from another registry location. Microsoft also noted it is not entirely file-less. Defenders could watch for mshta spawning from unusual parents and for new file-type handlers. The lesson: removal needs both the dropped files and the registry changes.
POSHSPY (documented 2017)
Mandiant described POSHSPY in April 2017 as a backdoor used by the group it tracks as APT29. It used WMI to store and persist its PowerShell-based code, and Mandiant saw it as a secondary backdoor. Microsoft cites it as a Type II example because the WMI repository is a shared data store rather than a plain file. Useful signals are new WMI event subscriptions and PowerShell started by WMI. The lesson: persistence in WMI is easy to miss unless you look for it.
Astaroth (2019 and 2020)
Astaroth is an information stealer, a form of spyware. Microsoft's 2019 analysis described a campaign that used only system tools: a shortcut file launched WMIC, then BITSAdmin, certutil and regsvr32, and the final stage was injected into a legitimate userinit.exe process. Microsoft noted that some decoded files were still written to disk, so "memory-only" overstates the case. A 2020 follow-up showed a revised chain that avoided WMIC and used techniques such as process hollowing. The lesson: a spike in WMIC use first exposed the campaign, so baselines pay off.
One caution: public reporting on memory-resident ransomware varies in quality. We have left it out rather than cite weak evidence.
Warning Signs and Indicators of Compromise
An indicator of compromise (IOC) is a clue that something may be wrong. No single clue proves a fileless infection. Weigh signals together and ask what changed.
Higher-value signals
Office, browser or email applications spawning PowerShell, mshta, wscript or regsvr32.
New WMI event filters, consumers or bindings that run commands.
Registry values holding long encoded strings, or odd file-type handlers.
Security tooling disabled or changed without an approved ticket.
Admin or remote-tool use from accounts or hosts that never did it before.
Medium-value signals
Scheduled tasks or services that launch interpreters.
Outbound connections from script hosts or from system processes that normally stay quiet.
Sudden privilege changes, new tokens or odd sign-in locations.
Lower-value signals on their own
Encoded PowerShell arguments, since some management tools use them legitimately.
Heavy use of built-in tools during maintenance windows.
Baselining is what turns these into useful alerts. Record which accounts, hosts and tools normally run scripts, then investigate departures from that picture. Monitoring gaps, such as missing command-line logging or short log retention, are warning signs in their own right.
How to Detect Fileless Malware
Detection is about behavior and context. Because the payload may leave little behind, you need telemetry that records what ran, who ran it, from where and what it talked to. Treat each signal as a lead, not a verdict.
Core telemetry sources
Process and command-line telemetry: Windows Security event 4688 can record process creation, and command-line capture needs its own policy setting turned on. Sysmon Event ID 1 adds the parent process, hashes and a process GUID for correlation.
PowerShell logging: Script block logging records the text of executed script blocks, including dynamically generated code, as event 4104, according to Microsoft's blue team guidance. That post also warns that some options are very noisy and advises forwarding logs off the machine quickly, since attackers may flood local logs. Module logging and transcription add more context. Check your own environment for the exact IDs and settings.
AMSI: The Antimalware Scan Interface lets applications pass content to the installed antimalware product. PowerShell, Windows Script Host, JavaScript, VBScript and Office VBA macros integrate with it, so obfuscated scripts can be inspected as they run.
Sysmon events: Event 3 (network connection) is disabled by default because of volume. Event 7 (image load) is high volume. Event 8 (CreateRemoteThread) is low volume and high signal for injection. Event 10 (process access) can be noisy, so filter it. Events 12 to 14 cover registry changes. Events 19, 20 and 21 record WMI filter, consumer and binding activity.
Network and DNS data: unusual destinations, rare domains and beaconing patterns from hosts that should be quiet. See this network security guide for the wider picture.
Identity logs: impossible travel, new devices, token abuse and privilege changes often reveal malware-free intrusion.
Memory and endpoint investigation
EDR tools build process trees and can flag suspicious chains. Memory scanning looks for injected or unbacked executable regions. MITRE's guidance on T1620 describes watching for memory allocation followed by execution inside one process, and for memory switched from writable to executable with no matching file on disk. Forensic memory capture helps when a host is already suspected, which links to the response section below.
Behavior to telemetry to triage
Behavior | Useful telemetry | Investigation question | Potential false positive |
Office app launches PowerShell | Process creation, command line, parent process | Did a user just open a document from email? | Approved automation or add-ins |
Encoded or obfuscated script | Script block logging, AMSI detections | What does the decoded content do? | Management or deployment tools |
New WMI event subscription | Sysmon 19 to 21, WMI operational logs | Who created it and why does it run a command? | Monitoring software and inventory tools |
Registry holds encoded payload or new handler | Sysmon 12 to 14, registry audit | Is the value read by a script interpreter? | Installers and software updates |
Code injected into a trusted process | Sysmon 8 and 10, EDR memory alerts | Is the source process expected to do this? | Security tools, debuggers, some game or accessibility software |
Script host contacts the internet | Network, DNS and proxy logs | Is the destination new or rare? | Legitimate updaters |
Sign-in plus unusual admin tool use | Identity and endpoint logs | Does the account normally do this from this host? | New employee or on-call work |
Detection engineering, hunting and validation
Map your detections to ATT&CK techniques so gaps are visible. Hunt on rare parent-child process pairs, uncommon command lines and script hosts with network activity. Keep logs long enough to reconstruct a slow intrusion, tune noisy rules and rank alerts by asset value. Detection logic based on machine learning can be tricked too, so keep adversarial AI risk in mind. Validate coverage safely with approved test cases in a lab, run by an authorized team, not with live malicious code. A short example of safe, illustrative logic: alert when a document-handling application starts a script interpreter, then enrich with the user, host and destination before escalating.
Best Fileless Malware Detection and Defense Solutions
No product category stops everything, and none replaces the others. Think of them as layers that see different things. Microsoft's documentation says Defender for Endpoint uses AMSI, behavior monitoring, memory scanning and boot sector protection, plus cloud machine learning, to inspect fileless threats. Other established vendors offer comparable capabilities, so confirm current features in each vendor's own documentation before buying.
Category | Best fit | What it sees | Main limits |
Next-gen antivirus (EPP) | Every endpoint as a baseline | Files, behavior, some memory and script content | Limited investigation and response depth |
EDR | Teams that can triage alerts | Process trees, memory, registry, scripts on endpoints | Needs skilled analysts; endpoint-only view |
XDR | Teams wanting correlation across email, identity, endpoint and cloud | Linked signals from several domains | Works best inside one vendor's ecosystem; integration effort |
MDR | Small teams without a 24/7 SOC | Outsourced monitoring and response on your telemetry | Depends on provider scope and your response authority |
SIEM | Organizations needing central logs and compliance | Logs from many sources, custom correlation | Needs tuning and staff; only as good as the logs fed in |
NDR | Networks with unmanaged or legacy devices | Traffic, lateral movement, beaconing | Limited endpoint detail; encrypted traffic |
Identity security | Cloud and hybrid environments | Sign-ins, tokens, privilege abuse | Does not see endpoint memory activity |
Application control | Controlled, well-managed endpoints | Blocks unapproved programs and scripts | Admin effort; does not stop trusted-tool abuse alone |
Costs vary by licensing, endpoint count and service level, so ask vendors for written pricing against your own scope. Prices are not published here because they change often and differ by region. If you need outside help running these layers, a guide to cybersecurity services explains the main service types. Teams running virtual desktops should also check how their tools work on desktop-as-a-service platforms.
How to Prevent Fileless Malware
Prevention works best as defense in depth. Order your work by risk, then separate controls by purpose.
Priority actions
Patch and manage vulnerabilities (prevent): close exploited holes in operating systems, browsers and Office first.
Use phishing-resistant MFA (prevent): passkeys or hardware keys reduce stolen-password risk.
Apply least privilege (prevent): remove standing local admin rights and limit remote administration tools.
Restrict scripting (prevent): block Office macros from the internet, and use application control and PowerShell Constrained Language Mode where suitable. Microsoft notes administrators can bypass Constrained Language Mode, and that it fails if users can write to trusted folders, so test it.
Enable attack surface and endpoint protections (prevent and detect): turn on tamper protection, behavior monitoring and AMSI-aware products.
Log centrally (detect): forward PowerShell, Sysmon, identity and network logs off the host.
Segment networks (contain): limit what a compromised endpoint can reach. See this network security guide for design ideas.
Back up and rehearse recovery (recover): keep offline copies and test restores.
Train people (prevent): use realistic exercises, such as phishing simulation software.
Build secure applications (prevent): review authentication, authorization and API design.
Run authorized security assessments (validate): confirm that fixes work.
Protect AI-enabled tools as well. Assistants and automation with command access need tight permissions, as discussed in this AI security overview.
![]() Turn a Long Fix List into a Ranked Plan Patching and hardening only help when you know which weaknesses matter most. A vulnerability assessment ranks known technical issues, and remediation verification retests fixes where agreed so you can confirm they hold.
Ask About Remediation Verification A penetration test does not guarantee security or compliance. |
What to Do If You Suspect a Fileless Attack
First, separate two situations. A business with a security team or provider should follow its incident response plan. An individual on a home computer can usually take simpler steps, described at the end of this section. NIST's incident response guidance, published in April 2025, is a trusted framework to build a plan around.
Enterprise response steps
Triage: confirm the alert, list affected hosts and accounts, and note times.
Escalate: involve the security lead, legal and, if needed, an incident response firm.
Contain proportionately: isolate the host from the network using your EDR or switch controls rather than powering it off, and disable suspect accounts.
Preserve volatile evidence: if your team is trained, capture memory and key logs first. Restarting or wiping a machine can destroy evidence of injected code and running processes, and can make scope harder to prove.
Investigate: review the process tree, script block logs, WMI subscriptions, registry autostart locations, scheduled tasks, and network and identity activity.
Determine scope: hunt for the same behavior on other hosts and accounts.
Eradicate: remove persistence, reset exposed credentials and tokens, and rebuild hosts when trust is lost.
Fix the root cause: close the entry point, patch and tighten logging.
Restore and monitor: return systems carefully and watch for reappearance.
Learn: write down what worked and what did not.
Home and small-office troubleshooting
Disconnect the device from the network, run a full scan with a reputable security product, change important passwords from a different clean device, and enable MFA. If you cannot explain what happened or the device holds sensitive data, ask a qualified professional. Easy removal is never guaranteed. Articsledge's testing services do not replace emergency incident response or malware removal.
Fileless Threats Across Windows, macOS, Linux and Cloud
Much of this guide describes Windows, where PowerShell, WMI and the registry give attackers many native tools. Other platforms are not immune, but the mechanics differ.
Windows: PowerShell, WMI, mshta, regsvr32, scheduled tasks and the registry are the usual instruments.
macOS: attackers can abuse shell scripts, AppleScript and signed utilities. MITRE lists PowerShell and reflective loading on macOS and Linux as well, but Windows registry and WMI guidance does not apply.
Linux: in-memory execution can happen through shell scripts and anonymous executable memory. MITRE's T1620 analytics describe memory marked executable with no ELF file on disk. Container hosts add exposure, so see cloud-native security and managed Kubernetes services.
Cloud and hybrid: the "land" being lived off is often the control plane. Stolen keys, tokens and over-privileged roles allow malicious actions with approved tools. Start with this cloud security guide, clarify who owns what in a public cloud model, and use cloud governance to keep permissions in check.
Visibility also differs by platform. Windows has rich event logging, while Linux and cloud workloads depend on auditd-style tools, agent telemetry and cloud audit logs. Do not copy Windows registry or PowerShell advice onto systems that lack them.
How to Choose the Right Defense Solution
Choose by your environment, not by a headline. Start by listing assets and operating systems, existing tools and staff skill, then score candidates against the same criteria.
Criterion | What to ask | Weight (1 to 5, your choice) |
Coverage | Which operating systems, servers and cloud workloads are supported? | Set by asset mix |
Detection quality | How does it handle script content, memory activity and LOLBin chains? | Often high |
Alert quality | How many alerts need human review in a trial? | High for small teams |
Response | Can it isolate hosts, kill processes and collect memory? | High |
Integration | Does it feed your SIEM, identity and ticketing tools? | Medium |
Data handling | Where is telemetry stored, and for how long? | Set by compliance |
Operations | Who tunes and runs it, and is MDR needed? | High for lean teams |
Total cost | Licenses, storage, services, training and staff time | Medium |
Run a proof of concept on real hosts with approved test cases. Check for false positives on your admin tools, ask for a written statement of what is and is not covered, and compare response times. Keep this evaluation separate from buying an external assessment, because a security product and a penetration test answer different questions. Also confirm how AI features inside products are governed, using AI security governance as a reference.
Where Penetration Testing Fits into Fileless Malware Defense
A penetration test is an authorized attempt to find and safely exploit weaknesses in a defined scope. A vulnerability assessment identifies and ranks known technical issues, usually with less hands-on exploitation. Both can reduce the number of ways in. Neither detects live fileless malware or replaces monitoring.
What testing can and cannot show
It can show: weak authentication, broken authorization, exposed APIs and misconfigurations that give attackers an initial foothold; and whether a fix actually worked.
It cannot show: whether your SOC would spot a script chain at 3 a.m., or guarantee that no weakness remains. Articsledge's own service page states that a penetration test does not guarantee security, does not establish compliance and cannot find every vulnerability.
How it differs from other services
Red teaming simulates a determined adversary against your detection and response, usually over a longer period, and is a different engagement. If your systems include AI features, AI red teaming, AI penetration testing and AI security testing cover model-specific risks. Continuous monitoring belongs to a SOC or MDR provider. Emergency incident response is a separate, urgent service.
Before you request testing
Define which applications, APIs or SaaS tools are in scope, who authorizes testing, and what documentation and test accounts you can supply. Articsledge offers web application, API and SaaS testing, vulnerability assessment and remediation verification, plus cloud configuration review within an agreed environment. Each engagement depends on technical scoping, qualified specialist availability, agreed commercial terms and written authorization. Our broader guide to penetration testing services explains what to ask any provider. Do not email passwords or production secrets in an initial inquiry. For general context, the main cybersecurity guide covers where testing fits.
![]() Decide What to Test First If you are weighing an external assessment, start with a scoping conversation: which applications, APIs or SaaS tools matter most, who must authorize testing, and what a useful report looks like for your team.
Discuss Your Security Assessment Timelines and pricing are set during scoping. |
Fileless Malware Security Checklist and 30/60/90-Day Plan
Use this plan to turn the guide into action. Small teams should finish the essentials first. Mature teams can then move to the advanced items.
Quick checklist
Command-line process logging and PowerShell script block logging are on and forwarded.
Macros from the internet are blocked and local admin rights are removed.
Phishing-resistant MFA covers email, VPN and admin accounts.
EDR or next-gen protection runs on every endpoint, with tamper protection on.
Backups are offline and a restore has been tested.
A written incident plan names who can isolate a host.
First 30 days
Turn on and centralize logs, patch internet-facing and Office software, remove unneeded admin rights and confirm endpoint protection coverage. Small organizations should consider an MDR provider if nobody can watch alerts after hours.
Days 31 to 60
Deploy Sysmon with a tuned configuration, build baselines for scripts and WMI, add application control in audit mode, segment sensitive networks and run a tabletop exercise based on a suspected fileless intrusion.
Days 61 to 90
Map detections to ATT&CK, begin regular threat hunts, enforce stricter script policies, test restore and response runbooks, and schedule an authorized assessment of exposed applications and APIs. Mature teams can add memory forensics practice and purple-team validation.
FAQ
What is fileless malware in simple terms?
It is an attack that runs code in memory or through trusted built-in tools instead of relying on a classic malicious program file. It often leaves fewer files, but it still leaves clues in logs, the registry and network traffic.
How does fileless malware work?
An attacker gets in through phishing, stolen credentials or a vulnerable service, then uses tools such as PowerShell or WMI to run code in memory. Some attacks add persistence in the registry or WMI so they can restart.
Is fileless malware the same as living-off-the-land activity?
No. LOTL means abusing legitimate tools and features. Fileless describes how code runs. They often overlap, and one can exist without the other.
Can antivirus detect fileless malware?
Often yes, in part. Modern products use AMSI, behavior monitoring and memory scanning. Legacy file-scanning alone is weak against it, and no single control catches everything.
Does fileless malware survive a reboot?
Some does not, because code held only in memory is lost on restart. Other threats persist through the registry, WMI subscriptions or scheduled tasks, so a reboot should never be treated as a cure.
What are some real examples?
Well-documented cases include Poweliks, Kovter, POSHSPY and Astaroth. Each mixes fileless behavior with some file or tool activity.
How does fileless malware use PowerShell?
PowerShell can run scripts from memory, download further stages and call system functions. Because administrators use it daily, defenders rely on script block logging, AMSI and context to tell good from bad.
How can organizations detect memory-based attacks?
Combine EDR memory scanning, process telemetry, Sysmon injection events and script logs. Watch for unusual parent-child processes and memory made executable inside processes that should not do that.
What are the warning signs of infection?
Look for Office apps starting script interpreters, new WMI subscriptions, odd registry values, disabled security tools and unusual admin activity. No single sign is proof.
Is fileless malware more dangerous than traditional malware?
Not automatically. It can be harder to find with file-based tools, but a strong logging and behavior-detection program can catch it. Impact depends on what the attacker reaches.
Can Macs and Linux systems experience fileless attacks?
Yes. Memory-based execution and script abuse exist on both, though Windows-specific tools like WMI and the registry do not apply. Visibility and defenses differ by platform.
What is the best defense against fileless malware?
There is no single best control. Layer patching, phishing-resistant MFA, least privilege, application control, script restrictions, centralized logging and EDR, then rehearse your response.
What is the difference between EDR and XDR?
EDR focuses on endpoint activity. XDR correlates signals from endpoints plus other sources such as email, identity and cloud, which can speed investigation but depends on integration quality.
Is Microsoft Defender sufficient for fileless malware protection?
Microsoft documents AMSI, behavior monitoring and memory scanning in Defender for Endpoint. Whether it is enough depends on licensing, configuration, monitoring skill and your other platforms, so test it against your own needs.
Can penetration testing help reduce fileless attack exposure?
It can reduce some entry-point weaknesses, such as flawed authentication or exposed APIs. It does not detect or stop fileless malware by itself, and it does not replace monitoring.
What should an organization do after a suspected fileless attack?
Follow your response plan: escalate, isolate the host, preserve volatile evidence, investigate scope, remove persistence, reset credentials and fix the entry point. Avoid reflexive reboots or wipes before evidence is captured.
Key Takeaways
Fileless malware is a family of techniques, not one clean category, and Microsoft itself says there is no single definition.
Fileless does not mean invisible. Behavior, logs, memory and identity data all offer detection chances.
Fileless, LOTL and malware-free describe different things. CrowdStrike's 82% malware-free figure is not a fileless statistic.
Persistence through the registry, WMI or scheduled tasks means a reboot is not a fix.
PowerShell script block logging, command-line process logging and Sysmon give the best evidence, but they need configuration and retention.
EDR, XDR, MDR, SIEM, NDR and identity security are complementary layers, not interchangeable products.
Do not reboot or wipe a suspected host before deciding what evidence to capture.
Penetration testing can reduce entry-point weaknesses but cannot detect or stop fileless malware on its own.
Baselines make alerts meaningful. Know what normal looks like before you hunt.
Actionable Next Steps
Confirm that command-line process logging and PowerShell script block logging are enabled and forwarded to a central place.
Check endpoint protection coverage and tamper protection on every device.
Block internet macros and remove unneeded local admin rights.
Roll out phishing-resistant MFA for email, remote access and admins.
Build a short list of normal script, WMI and remote-tool use to serve as a baseline.
Write and rehearse an incident plan that covers volatile evidence.
Decide whether you need MDR for after-hours coverage.
If you run web applications, APIs or SaaS tools, request a scoped assessment and ask for retesting of fixes.
Glossary
Fileless malware: malicious activity that runs in memory or through trusted tools with little or no new executable file.
Memory-resident code: code that runs from RAM and may disappear after a restart unless something restarts it.
LOTL / LOLBin: living off the land means using legitimate tools for malicious aims. A LOLBin is such a program.
PowerShell: a Windows command-line shell and scripting language.
WMI: Windows Management Instrumentation, a framework for managing systems locally and remotely.
AMSI: the Antimalware Scan Interface, which lets apps send content to antimalware tools for inspection.
EDR: endpoint detection and response, which records and responds to endpoint activity.
XDR: extended detection and response, which correlates signals across several security domains.
MDR: managed detection and response, an outsourced monitoring and response service.
SIEM: security information and event management, a platform for collecting and correlating logs.
IOC: indicator of compromise, a clue that an intrusion may have occurred.
TTP: tactics, techniques and procedures, the patterns of how attackers operate.
Process injection: running code inside another process.
Persistence: a method that keeps an attacker's access after a restart or logoff.
Behavioral detection: finding threats from what software does, not from what a file looks like.
Penetration testing: authorized testing that tries to find and safely exploit weaknesses in an agreed scope.
Sources & References
Fileless threats. Microsoft Learn, last updated 2024-04-24.
Sysmon events. Microsoft Learn, last updated 2026-02-24.
PowerShell the Blue Team. Microsoft PowerShell Team blog, 2015-06-09.
Antimalware Scan Interface (AMSI) portal. Microsoft Learn, last updated 2025-07-10.
T1059.001 PowerShell. MITRE ATT&CK, modified 2026-09-18.
T1047 Windows Management Instrumentation. MITRE ATT&CK, modified 2026-11-03.
T1055 Process Injection. MITRE ATT&CK, modified 2026-09-11.
T1620 Reflective Code Loading. MITRE ATT&CK, modified 2026-05-12.
T1112 Modify Registry. MITRE ATT&CK, modified 2026-09-11.
Dissecting One of APT29's Fileless WMI and PowerShell Backdoors (POSHSPY). Mandiant, 2017-04-03.
Poweliks: the persistent malware without a file. G Data, 2014-07-31.
Kovter analysis. Microsoft Security Blog, n.d.
Dismantling a fileless campaign: Astaroth attack. Microsoft Security Blog, 2019-07-08.
Latest Astaroth living-off-the-land attacks are even more invisible but not less observable. Microsoft Security Blog, 2020-03-23.
CrowdStrike 2026 Global Threat Report: The Evasive Adversary Wields AI. CrowdStrike, 2026-02-24.
Identifying and mitigating living off the land techniques. Australian Cyber Security Centre and partner agencies, n.d.
NIST SP 800-61 Rev. 3: Incident Response Recommendations and Considerations. NIST, April 2025.






