⚔️ Cyber Kill Chain Guide - Anatomy of Attack
Seven chapters on modelling of attacks in international practice: Lockheed Martin Kill Chain (Hutchins, Cloppert, Amin 2011), MITRE ATT&CK 14 tactics, Unified Kill Chain (Paul Pauls), Diamond Model, D3FEND defense matrix, traces in Windows, macOS and Linux magazines and related to reporting of Article 23 of the NIS2 incidents. Examples are fictional, designed for training.
Basis for attack modelling
Kill Chain principle - breaking one stage, mission fails
Lockheed Martin Cyber Kill Chain - Hutchins, Cloppert, Amin (2011)
- 1. IntelligenceReconnaissanceOSINT, scanning
- 2. ArmamentWeaponizationexploit + payload
- 3. DeliveryDeliveryphishing, USB, web
- 4. OperationExploitationvulnerability or user click
- 5. InstallationInstallationbackdoor, durability
- 6. Management (C2)Command & Controlremote control
- 7. ObjectivesActions on Objectivesdata export, encryption
The defender can interrupt the chain at any stage. Six Actions of Action: To detect, deny, dismantle, trouble, deceive, destroy. The earlier the chain is interrupted, the less damage. The attacker must succeed in all seven phases, the defender must stop one.
History
The Cyber Kill Chain model was formalized in 2011 by Lockheed Martin researchers Hutchins, Cloppert and Amin. Model adapted from military 'F2T2EA' kit chain (Find, Fix, Track, Target, Engage, Assess) in the US Air Force.
Basic principle
The attack is a sequential chain - every next phase needs a previous result. By stopping one stage, the whole attack fails. This means that the defenders do not have to be perfect in all phases - it is enough to break one.
In practice
Three modeling, threat intelligence (TI), detection engineering, incident response plans, security gap analysis, purple team simulations.
Restrictions
Model created for APT/network intrusion analysis. Less suited for internal threats, supply chain attacks, cloud-native attacks and compromised SaaS sessions where 'supply' and 'operation' are foggy.
Lockheed Martin 7 Phases
From intelligence to objectives - each phase with protection
| Phase | Attacker | Defender |
|---|---|---|
| 1. IntelligenceReconnaissance | Attacker
| Defender
|
| 2. ArmamentWeaponization | Attacker
| Defender
|
| 3. DeliveryDelivery | Attacker
| Defender
|
| 4. OperationExploitation | Attacker
| Defender
|
| 5. InstallationInstallation | Attacker
| Defender
|
| 6. Management (C2)Command & Control | Attacker
| Defender
|
| 7. ObjectivesActions on Objectives | Attacker
| Defender
|
A step-by-step attack - and where it can be stopped
Seven phases in one example. The Green Statement shows what the defender can do precisely in this step.
Attack with phishing email
- 1Attacker1. Intelligence: collects employees' names and emails from the site and LinkedIn
The attacker does not send anything yet - he learns about an organization from public sources.
- 2Attacker2. Gun preparation: document with malware macro
Prepare an attachment that will trigger the code when opening.
- 3Attacker → Organisation3. Delivery: e-mail'Invoice.docm '
Defender can: e-mail filter, SPF/DKIM/DMARC, employee recognizes and reports
This is the first time an attacker touches the organization - and the first chance to break the chain.
- 4Organisation4. Use: employee opens attachment, macro starts
Defender: macro from the Internet banned, EDR stops suspicious process.
- 5Organisation5. Install: malware set to start after restart
Defender: Application control, automatic starting point monitoring.
- 6Organisation → Attacker6. Management (C2): the computer communicates with the attacker's server
Defender can: DNS filter, output traffic restriction, network monitoring (NDR)
The infected computer itself calls the attacker - therefore, the limited outgoing traffic is a powerful protection.
- 7Attacker → Organisation7. Achievement of objectives: data collection, export or encryption
Defender can: segmentation, DLP, offline copies
If the chain is not interrupted earlier, here are the biggest losses.
- 8NoteThe earlier the chain is interrupted, the less damage.
The attacker must succeed in every step, the defender must stop one.
Typical commands in each phase
The attacker lines are for recognition and authorised testing, with no ready-made payloads. Use only on your own systems or with written permission.
1. Reconnaissance
-
Attacker: registration data and mail servers.
whois <domēns>; dig +noall +answer <domēns> MX -
Attacker: open ports and versions.
nmap -sV -Pn <mērķis> -
Defender: addresses with the most requests.
awk '{print $1}' access.log | sort | uniq -c | sort -rn | head
2. Weaponization
-
Defender: file identity and real type. Weaponization itself happens on the attacker's side and is not in your logs.
sha256sum <fails>; file <fails> -
Defender: macros in a document (oletools).
olevba <dokuments.docm> -
Defender: searches files for known indicators.
yara -r noteikumi.yar <mape>
3. Delivery
- GoPhish Application for authorised phishing training in an organisation.
-
Defender: the message route and SPF, DKIM, DMARC results.
grep -iE '^(received|authentication-results|return-path):' vestule.eml -
Defender: where a link leads without opening it in a browser.
curl -sIL <saite>
4. Exploitation
-
Authorised test: known vulnerabilities on your own system.
nuclei -u https://<savs-mērķis> -severity critical,high -
Defender: service errors around the time of the attack.
journalctl -p err --since '1 hour ago' -
Defender on Windows: services that stopped unexpectedly.
Get-WinEvent -FilterHashtable @{LogName='System';Id=7031,7034}
5. Installation
-
Defender on Linux: what starts automatically.
systemctl list-unit-files --state=enabled; crontab -l -
Defender on Windows: active scheduled tasks.
Get-ScheduledTask | Where-Object State -ne Disabled -
Defender: recently changed files.
find /etc /usr/local /tmp -type f -mtime -2
6. Command and control
-
Defender on Linux: active connections and processes.
ss -tunap state established -
Defender on Windows: the same.
Get-NetTCPConnection -State Established | Select-Object RemoteAddress,RemotePort,OwningProcess -
Defender: unusual DNS volume can be a control channel.
sudo tcpdump -nn -i any 'port 53' -c 100
7. Actions on objectives
-
Attacker on Windows: typical discovery after a break-in. Look for this pattern in logs.
whoami /all; net group 'Domain Admins' /domain -
Attacker: a common data exfiltration tool. Unexpected rclone on a server is an indicator.
rclone copy <mape> <attālināta-krātuve>: -
Defender: auditd records of sensitive file use.
ausearch -k <atslēga> -ts recent -
Defender on Windows: access to audited objects.
Get-WinEvent -FilterHashtable @{LogName='Security';Id=4663} -MaxEvents 50
In practice
This table serves as a detection engineering 'card' - each phase needs to know what artifact or signal indicates how it is going, so that SOC can establish the appropriate rule.
Reference
Lockheed Martin "Cyber Kill Chain' (2011), Hutchins, Cloppert, Amin - "Intelligence-Driven Computer Network Defense Informed by Analysis of Additional Campaigns and Intrusion Kill Chains'.
Lacmus test
Can your SOC reveal at least one artifact in each of the seven phases? If any phase is 'empty' - there's a blind strip that an attacker can use without respect.
MITRE ATT&CK - tactics, techniques, procedures
14 tactics - detailed fascinating behavior classification
MITRE ATT&CK Enterprise - 14 tactics in matrix order, each with three examples of techniques
- TA0043IntelligenceReconnaissanceKill Chain Phase: 1
- Active Scanning
- Gather Victim Identity Information
- Search Open Websites/Domains
- TA0042Resource PreparationResource DevelopmentKill Chain Phase: 2
- Acquire Infrastructure
- Compromise Accounts
- Develop Capabilities
- TA0001Initial accessInitial AccessKill Chain Phase: 3
- Phishing
- Drive-by Compromise
- Valid Accounts
- TA0002EnforcementExecutionKill Chain Phase: 4
- Command and Scripting Interpreter
- User Execution
- Scheduled Task/Job
- TA0003StabilityPersistenceKill Chain Phase: 5
- Boot or Logon Autostart Execution
- Create Account
- Scheduled Task/Job
- TA0004EnforcementPrivilege EscalationKill Chain phase: 4-7
- Abuse Elevation Control Mechanism
- Access Token Manipulation
- Exploitation for Privilege Escalation
- TA0005Circulation of protectionDefense EvasionKill Chain phase: 4-7
- Obfuscated Files or Information
- Indicator Removal
- Impair Defenses
- TA0006Obtaining access dataCredential AccessKill Chain Phase: 7
- Brute Force
- OS Credential Dumping
- Steal Web Session Cookie
- TA0007Environmental researchDiscoveryKill Chain Phase: 7
- Account Discovery
- Network Service Discovery
- Process Discovery
- TA0008NetworkingLateral MovementKill Chain Phase: 7
- Remote Services
- Internal Spearphishing
- Lateral Tool Transfer
- TA0009Data collectionCollectionKill Chain Phase: 7
- Email Collection
- Data from Local System
- Screen Capture
- TA0011Management (C2)Command and ControlKill Chain Phase: 6
- Application Layer Protocol
- Encrypted Channel
- Ingress Tool Transfer
- TA0010Data removalExfiltrationKill Chain Phase: 7
- Exfiltration Over C2 Channel
- Exfiltration Over Web Service
- Scheduled Transfer
- TA0040ImpactImpactKill Chain Phase: 7
- Data Encrypted for Impact
- Data Destruction
- Service Stop
Kill Chain phase is approximate conformity. ATT&CK's tactics are not consecutive: they are repeated by the attacker and are bypassing protection throughout the attack, so the increase of rights and the circumvention of protection are consistent with several phases.
From abstract to specific
- 1. Tactics - WhyTacticattacker's goal in this step, 14 tactics
- 2. Technical - asTechniquegeneral type, around 200 techniques
- 3. Sub-equipmentSub-techniquemore specific option, more than 400
- 4. ProcedureProcedurewhat a particular group did
Example: TA0002 Execution → T1059 Command and Scripting Interpreter → T1059.001 PowerShell → team specific.
Difference from LM
The LM Kill Chain has a linear chain (after → how). ATT&CK is a granular tactical catalogue (which → as → specifically). The attacker can stop, return, repeat phases - so 14 tactics are not necessarily consecutive.
Reference
MITRE ATT&CK Enterprise (attack.mitre.org). Established in 2013, publicly available since 2015, new versions are released approximately twice a year.
Practical use
ATT&CK Navigator - map which techniques your SOC covers with detection. Empty cells = blind strips. Purple team simulates procedures to verify detection.
Volume trap
Covering all ~400 sub-technics is a marathon work. Start from APT groups that target your sector (e.g. FIN7 for finance, Lazarus for energy) rather than all at once.
Unified Kill Chain - Paul Paul (2017)
3 phases of 18 phases - merging LM + ATT&CK
Unified Kill Chain - Paul Paul (2017, updated 2022)
A. Initial access In - Initial Foothold
The attacker gets the first stop.
- 1. IntelligenceReconnaissanceOSINT, scanning
- 2. Resource PreparationResource Developmentinfrastructure, tools
- 3. DeliveryDeliveryphishing, USB
- 4. Social engineeringSocial Engineeringmanipulation
- 5. OperationExploitationvulnerability, macro
- 6. StabilityPersistenceautomatic startup
- 7. Circulation of protectionDefense Evasionskiing
- 8. Management (C2)Command & Controlcontact with the attacker
B. Networking Through - Network Propagation
The attacker extends access and moves towards the target.
- 9. ConversionPivotingproxy, tunnel
- 10. Environmental researchDiscoveryAD, Network Card
- 11. EnforcementPrivilege EscalationUAC, sudo, token
- 12. EnforcementExecutionPowerShell, WMI
- 13. Obtaining access dataCredential AccessLASS, password mixing values
- 14. NetworkingLateral MovementRDP, SMB
C. Achievement of objectives Out - Action on Objectives
The attacker reaches why came: data, extortion or sabotage.
- 15. Data collectionCollectionpreparation, compression
- 16. Data removalExfiltrationcloud, DNA, e-mail
- 17. ImpactImpactencryption, destruction, DoS
- 18. Achievement of objectivesObjectivesstrategic result
The stages are repeated: after each movement in the network, the attacker often starts again with research and empowerment.
Why 18 phases?
LM 7 phases are too simplified for a modern attack - especially 'after access' activities. UKC divides them into stage B and clearly separates social engineering as a separate phase.
Reference
Paul Pauls - Unified Kill Chain (Master's Work, Cyber Security Academy, 2017; updated version 2022), unifiedkillchain.com.
In practice
For analysis of modern APT/ransomware operations - where the attacker returns several times between stages A and B (e.g. executes the 2.3, 4. intelligence iteration after each lateral movement).
Complexity
18 phases are more difficult to communicate to management than LM 7 phases. UKC is better suited for technical analysis, LM for presentations and risk communication.
Protection strategy
Courses of Action matrix + Diamond Model + MITRE D3FEND
Courses of Action (CoA) - six actions in each of the seven phases (Hutchins, Cloppert, Amin 2011)
| Phase | DiscoverDetect | DenyDeny | DisruptDisrupt | BurdenDegrade | DecideDeceive | DestroyDestroy |
|---|---|---|---|---|---|---|
| 1. IntelligenceReconnaissance | DiscoverWeb Log Analysis, Honeypot | DenyLess public information | DisruptRequest limits, CAPTCHA | BurdenTarpit - a slow answer to the scanner | DecideFalse DNA recordings, mammalports | DestroyNot legally allowed |
| 2. ArmamentWeaponization | DiscoverYARA rules, threat intelligence | Denyoutside the defender's network | Disruptoutside the defender's network | Burdenoutside the defender's network | DecideHoneyfiles, canary tokens | Destroyoutside the defender's network |
| 3. DeliveryDelivery | DiscoverAttachment sandbox, URL reputation | DenySPF/DKIM/DMRAC, USB ban | DisruptVerification of annexes before delivery | BurdenGreylisting, shipment limitation | DecideMunicipal Email Accounts | DestroyAbuse URL removal with CSIRT |
| 4. OperationExploitation | DiscoverEDR signals for operation | DenyPatches, DEP/ASLR, sandbox | DisruptExploit Guard, EDR blocks | BurdenSandbox Resource Limits | DecideMental services | DestroyDelete File, Change Passwords |
| 5. InstallationInstallation | DiscoverFIM, author supervision | DenyList of applications authorised, code signature | DisruptQuarantine, EDR roll-out | BurdenRight to enforce only temporarily (JIT) | DecideMotherbackdoor, fake register | DestroyReinstall from a pure image |
| 6. Management (C2)Command & Control | DiscoverDNA and DGA analysis, NID, JA3/JA4 | DenyOutgoing traffic filtering, IoC blocking | DisruptDNA Syncole, Proxy Block | BurdenZone width limits | DecideC2 channelling to honeynet | DestroyC2 single with law enforcement |
| 7. ObjectivesActions on Objectives | DiscoverDLP, Behavioural Analysis (UEBA) | DenyDLP blocking, segmentation | DisruptNetwork isolation of computer | BurdenLimit on the amount of data to be exported | DecideMental files, data traps | DestroyNetwork cut-off, renewal |
Diamond Model of Intrusion Analysis - Caltagirone, Pendergast, Betz (2013)
Each incident is described by four corners. Finding one can find the others.
Finding one corner:
- the victim reaches the capacity used;
- from capability to infrastructure;
- from infrastructure to other victims;
- all together allow the detection of the opponent (the award).
MITRE D3FEND 1.0 - protective tactics:
- modelling and strengthening (Model, Harden);
- detect and isolate (Detect, Isolate);
- deceive;
- expel and restore (Evict, Restore).
Courses of Action
6 types of action have no choice - an advocacy strategy ideally covers several stages so that the attacker has several potentially cut-off points.
Diamond Model
Threat intel analysis framework - each incident as a 4-hour ratio. When discovering one corner, you can find the rest. Perfect for tracking campaigns.
D3FEND
MITRE D3FEND (2021) - catalogue of defence activities according to ATT&CK techniques. D3fend.mitre.org - shows for each ATT&CK technique which D3FEND operations neutralise it.
Legal notice
In the 'Destroy' phase carefully - 'hack back' (attack back) is mostly illegal. Allowed: takeown with law enforcement, C2 synkhole, cooperation with CSIRT.
Detection of Windows, macOS and Linux
What the attacker leaves in the system and how to see it
An attack from the supply phase leaves traces in the host: entries in the magazine, new service, planned task. These are places where the chain can be seen and interrupted. The hardness (to be ruled out above) is surface guide.
The bottom row is where to watch - commands below. Figures are Windows event IDs, codes - MITRE ATT&CK techniques.
1. Logs - see and follow in real time
Windows
Recent safety events:
Get-WinEvent -LogName Security -MaxEvents 20 |
Format-Table TimeCreated, Id -AutoSize
Follow in real time (new every 5 s):
$since = Get-Date
while ($true) {
Start-Sleep 5; $now = Get-Date
Get-WinEvent -FilterHashtable @{
LogName='Security'; StartTime=$since
} -ErrorAction SilentlyContinue |
Format-Table TimeCreated, Id -AutoSize
$since = $now
}
Key IDs: 4624/4625 login, 4688 new process (to be turned on in audit policy), 4697 and 7045 new service, 4698 planned task, 1102 magazine deleted.
macOS
Last hour, failed actions:
log show --last 1h --style compact \
--predicate 'eventMessage CONTAINS[c] "fail"'
Follow in real time (e.g. SSH):
log stream --style compact \
--predicate 'process == "sshd"'
Linux
Last hour warnings:
journalctl --since "1 hour ago" -p warning
Follow in real time:
journalctl -f
sudo tail -F /var/log/auth.log # Debian/Ubuntu
sudo tail -F /var/log/secure # RHEL/Fedora
2. Services - Find new and stop
Windows
Operating services and recently installed:
Get-Service | Where-Object Status -eq Running
Get-WinEvent -FilterHashtable @{
LogName='System'; Id=7045 } -MaxEvents 10
Stop and disable
Stop-Service <serviss>
Set-Service <serviss> -StartupType Disabled
Revert
Set-Service <serviss> -StartupType Manual
Start-Service <serviss>
macOS
Loaded and third party launchd entries:
launchctl list | grep -v com.apple
ls /Library/LaunchDaemons /Library/LaunchAgents \
~/Library/LaunchAgents
Stop and disable
sudo launchctl bootout system \
/Library/LaunchDaemons/<label>.plist
sudo launchctl disable system/<label>
Revert
sudo launchctl enable system/<label>
sudo launchctl bootstrap system \
/Library/LaunchDaemons/<label>.plist
Linux
Operating and on-board services:
systemctl list-units --type=service --state=running
systemctl list-unit-files --state=enabled
Stop and disable
sudo systemctl disable --now <serviss>
Revert
sudo systemctl enable --now <serviss>
3. Planned tasks - review and set your own
Windows
Active tasks (the attacker hides resilience right here):
Get-ScheduledTask | Where-Object State -ne Disabled |
Select-Object TaskPath, TaskName, State
Set Daily Check
$a = New-ScheduledTaskAction -Execute powershell.exe `
-Argument '-NoProfile -File C:\Scripts\check.ps1'
$t = New-ScheduledTaskTrigger -Daily -At 06:00
Register-ScheduledTask -TaskName DailyCheck `
-Action $a -Trigger $t -User 'NT AUTHORITY\SYSTEM'
Delete
Unregister-ScheduledTask -TaskName DailyCheck -Confirm:$false
macOS
User cron and third party launchd:
crontab -l
launchctl list | grep -v com.apple
Set Daily Check
(crontab -l 2>/dev/null
echo '0 6 * * * /usr/local/bin/check.sh') | crontab -
Delete
crontab -l | grep -v /usr/local/bin/check.sh | crontab -
Linux
Timers and cron:
systemctl list-timers --all
crontab -l; ls /etc/cron.d /etc/cron.daily
Set Daily Check
sudo systemd-run --unit=daily-check \
--on-calendar='*-*-* 06:00:00' /usr/local/bin/check.sh
Delete
sudo systemctl stop daily-check.timer
NIST SP 800-92
Log Management: what to collect, how long to store, how to protect against deletion.
CIS Controls v8.1 - 8
Audit log management: centralised collection, time synchronization, regular review.
MITRE ATT&CK
T1543 service, T1053 planned task, T1070 foot quench - permanent and skiing techniques that are seen right here.
In practice
Logs should be sent to a central system (SIEM) - an attacker with administrator rights can delete the local log.
Article 23 NIS2 and national legislation
Incident reporting - 24 h / 72 h / 1 month after notification
Article 23 of Directive (EU) 2022/2555, transposed by each Member State into national law. Time limits shall count from the moment the entity becomes aware of a significant incident.
- 24 h Early warningEarly Warning, 23(4)(a)suspected abuse and cross-border effects
- 72 hours Incident notificationIncident Notification, 23(4)(b)initial assessment: severity, effect, IoC, if any
- on request Intermediate reportIntermediate Report, 23(4)(c)if the CSIRTs or the supervisory authority so requests
- 1 month Final reportFinal Report, 23(4)(d)within one month of notification: cause, measures, impact
If the incident is still ongoing within the deadline of the final report, a progress report and a final report shall be submitted within one month of the end of the incident.
Supervision - Competent authorities
Monitors relevant and important entities, may issue orders and impose penalties, check self-assessments and audits, cooperate with ENISA and other national authorities.
CSIRTs - Incident Prevention
Member State Command: Receives messages, assists in response, disseminates IoC and alerts, coordinates the detection of vulnerabilities.
Who should report?
The report shall be received by the CSIRT or by the competent authority in accordance with national law. Which institution and to which supervisory authority to report may also depend on the sector of the entity - a check before, not during, the incident.
Reference
Article 23 of Directive (EU) 2022/2555 and its transposing national legislation.
Practical link with Kill Chain
An important incident is defined by the impact, not by the phase: serious disruption of the service, financial loss or damage to others (Article 23 (3)). The deadline starts when you learn about an incident - the earlier you notice it in the chain, the more time you react.
Penalties
NIS2 provides penalties of up to 10 million euros or 2% of the annual turnover in the world for significant entities and up to 7 million euros or 1.4% for important entities. Non-communication itself is an offence - even if an incident was caused by a failure, not an attack.
Abbreviations
Abbreviations used in the Roadmap: the original name and what it means.
- APT
- Advanced Persistent Threat - a well-funded group that long and targeted attacks one organization.
- TTP
- Tactics, Techniques and Procedures - the course of action of the attacker.
- MITRE ATT&CK
- An attacker tactics (TA00xx) and a technical knowledge base.
- UKC
- Unified Kill Chain - Pola Pulsa model, which combines Kill Chain and ATT&CK in 18 phases.
- CoA
- Courses of Action - types of actions of the defender: detect, deny, interfere, weaken, deceive, destroy.
- F2T2EA
- Find, Fix, Track, Target, Engage, Assess - the military target chain model from which Kill Chain originated.
- D3FEND
- MITRE knowledge base for defence techniques - ATT&CK opposite.
- OSINT
- Open-Source Intelligence - collecting information from public sources.
- IoC
- Indicator of Compromise - a sign that the system was cracked.
- C2
- Command and Control - the control channel of the attacker to an infected computer.
- RAT
- Remote Access Trojan - malware that gives the attacker remote access.
- DGA
- Domain Generation Algorithm - malware creates new domains every day to make it harder to block.
- LSASS
- Local Security Authority Subsystem Service - Windows process, memory of which is login data - often target of attack.
- UAC
- User Account Control - Windows approval window for enhanced rights.
- WMI
- Windows Management Instrumentation - Windows management interface, which is also used by attackers.
- SMB
- Server Message Block - Windows file sharing protocol.
- ASLR / DEP
- Address Space Layout Randomization / Data Execution Prevention - memory protection which makes it difficult to use vulnerabilities.
- SPF / DKIM / DMARC
- Checks on the sender of e-mails which make it difficult for counterfeit e-mails.
- EDR / NDR
- Endpoint / Network Detection and Response - monitoring and responding to computers/networks.
- NIDS
- Network Intrusion Detection System - discovery of intrusion into the network stream.
- UEBA
- User and Entity Behavior Analytics - detection of behavioural abnormalities.
- FIM
- File Integrity Monitoring - warns if important files change.
- SIEM
- Security Information and Event Management - central analysis of magazines.
- YARA
- Rule language for the recognition of malware in files.
- JA3 / JA4
- TLS connection imprint for recognising tool or malware.
- KEV
- Known Exploited Vulnerabilities - CISA catalog with already used injuries.
- CVE
- Common Vulnerabilities and Exposures - vulnerability identifier (CVE-year-number).
- CSIRT
- Computer Security Incident Response Team - incident response team.
- ASM
- Attack Surface Management - a permanent inventory of their systems on the internet.
- JIT
- Just-In-Time - right to a short period of time only.