Key Takeaways
- The cyber kill chain splits an intrusion into seven stages, from reconnaissance to actions on objectives. Each stage depends on the one before it.
- Stopping a single stage stops that attack path. The attacker has to win all seven; the defender needs to win one.
- Defense in depth puts controls at several stages, so a miss at one link gets caught at the next.
- MITRE ATT&CK catalogs the techniques used inside each stage. The kill chain gives them an order. Most mature teams use both.
- Reachability data shows which link is cheapest to break in your environment. It is often a firewall rule or a segmentation change, not a patch.
An attacker who lands a phishing payload but can't call home has nothing. An attacker who gets a shell on a web server but can't reach the database behind it never touches the data. Every intrusion is a sequence, and every step in that sequence has to work. The cyber kill chain turns that dependency into a defensive strategy: find the link you can break, and break it.
The Stages of the Cyber Kill Chain
Lockheed Martin published the model in 2011, adapting a military targeting concept to network intrusions. It has seven stages.
1. Reconnaissance
The attacker studies the target. They pull employee names from LinkedIn, enumerate subdomains, and scan for internet-facing services. Example: finding a VPN appliance running a firmware version with a public CVE.
2. Weaponization
The attacker pairs an exploit with a payload. This happens on their own infrastructure, so defenders never see it. Example: building a malicious document that drops a loader when opened.
3. Delivery
The weapon reaches the target. Email attachments, malicious links, exposed services, and compromised suppliers are the common routes. Example: a spear-phishing email sent to a finance analyst.
4. Exploitation
The vulnerability fires. Code runs on a host the attacker doesn't own. Example: a deserialization flaw in an exposed application server grants remote code execution.
5. Installation
The attacker sets up persistence so access survives a reboot or a password reset. Example: a scheduled task that relaunches a backdoor every hour.
6. Command and control
The compromised host opens a channel back to the attacker. Command and control (C2) traffic often hides in HTTPS or DNS to look like normal browsing. Example: a beacon that checks in with a cloud-hosted server every few minutes.
7. Actions on objectives
The attacker does what they came to do: steal data, deploy ransomware, or move laterally toward a more valuable system. Example: staging and exfiltrating a customer database.
Why Disrupting a Single Stage Halts the Attack
The seven stages form a dependency chain. Exploitation needs successful delivery. Installation needs successful exploitation. Command and control needs a running implant. If any stage fails, every stage after it has nothing to build on.
That flips the usual framing of security as a game the defender has to win everywhere. On any single attack path, the attacker has to succeed seven times in a row. The defender needs one success.
Consider an attacker who exploits a vulnerable server in the DMZ. If that server has no route to the internal network, the attacker has a foothold with nowhere to go. If outbound traffic from the DMZ is restricted to known destinations, the implant can't reach its C2 server. Either control ends that intrusion, even though exploitation succeeded.
Breaking a link does not make the attacker disappear. A determined adversary will look for another path. But they have to restart from an earlier stage, which costs time and creates more chances to be detected. Breaking links early is cheapest. Breaking them anywhere is worth it.
Defensive Controls Mapped to Each Kill Chain Stage
Defense in depth means each stage meets at least one control, so a gap at one stage doesn't decide the outcome.
Two things stand out in this table. Network controls (segmentation, firewall rules, egress filtering) show up at several stages. And the controls most teams lean on hardest, patching and EDR, cover only one or two stages each.
Mapping which paths actually exist through those controls is the job of attack path analysis. For a comparison of how vendors approach it, see our guide to the best attack path analysis tools.
How the Cyber Kill Chain Relates to MITRE ATT&CK
The two frameworks answer different questions. The cyber kill chain answers "what order does an attack happen in?" MITRE ATT&CK answers "what specific techniques do attackers use?"
ATT&CK is a large, continuously updated matrix of tactics and techniques drawn from real intrusions. Its tactics cover ground the kill chain compresses into a single stage. Privilege escalation, credential access, discovery, and lateral movement all happen after the initial compromise, and the original kill chain doesn't break them out.
In practice, teams use each for different work:
- Kill chain: explaining an incident to leadership, planning where to invest in controls, and structuring incident response.
- ATT&CK: writing detection rules, measuring detection coverage, and running purple team exercises.
The unified kill chain bridges the two. It keeps the kill chain's sequence and adds the post-compromise phases ATT&CK documents.
Finding the Easiest Link to Break in Your Environment
Knowing the stages is the easy part. The harder question is which link to break first in your network, with the people and change windows you actually have.
Most programs default to the exploitation stage: find the vulnerability, patch it. That works until the backlog has tens of thousands of "critical" findings and the change window is weeks away. Frontier AI models are also shrinking the time from CVE disclosure to working exploit, so ranking by CVSS or presumed exploitability no longer buys much safety.
Reachability changes the question. Instead of asking how severe a vulnerability is, it asks whether an attacker can get to it from a realistic foothold: the internet, a third-party connection, or a compromised internal segment. The answer is yes or no. Either a network path exists to the vulnerable service, or it doesn't.
That answer maps directly onto the kill chain. A vulnerability with no reachable path can't be exploited, so the chain is already broken at that stage. In most environments, about 1% of vulnerabilities are reachable and exploitable. Those are the links an attacker can actually use.
For each reachable one, the full attack path shows every hop: the firewall that allows the traffic, the NAT rule, the route, the listening service. Each hop is a place to break the chain. Sometimes the fastest fix is the patch. Often it's a single firewall rule, a segment change, or a compensating control that IT can ship this week without touching production systems.
This is the approach Astelia is built on. Astelia maps real network topology through read-only integrations with the your security stack. Its agentic AI works out what each exploit requires and checks those requirements against the topology. The output is the short list of reachable vulnerabilities, the attack path behind each, and multiple remediation and remediation options. For more on why reachability is missing from most exposure management programs, read the blind spot in exposure management.
See which links in your kill chains are reachable today: Book a demo.
FAQs
Who created the cyber kill chain model?
Eric Hutchins, Michael Cloppert, and Rohan Amin, analysts at Lockheed Martin, introduced it in a 2011 paper titled "Intelligence-Driven Computer Network Defense Informed by Analysis of Adversary Campaigns and Intrusion Kill Chains." They adapted it from the military F2T2EA targeting cycle (find, fix, track, target, engage, assess). Their core idea was that defenders could study a full intrusion campaign, not just the single alert in front of them, and use that pattern to anticipate the next one.
Is the cyber kill chain still relevant today?
Yes, with limits. Critics note that it centers on malware and the network perimeter. Many current attacks skip weaponization and delivery entirely, for example logging in with stolen credentials or abusing a misconfigured SaaS tenant. The model is still valuable for sequencing an attack, briefing executives, and deciding where controls are thin. Pair it with ATT&CK for technique-level detail and with reachability data for prioritization.
How does the unified kill chain differ?
Paul Pols developed the unified kill chain in 2017, combining the Lockheed Martin model with MITRE ATT&CK. It has 18 phases grouped into three cycles: In (gaining an initial foothold), Through (propagating across the network), and Out (acting on objectives). The key difference is that it treats attacks as loops. After a foothold, an attacker runs reconnaissance and exploitation again from the inside, which the original linear model doesn't capture.
Does the kill chain apply to insider threats?
Only partly. A malicious insider already has access, so reconnaissance, weaponization, delivery, and often exploitation don't apply. They usually start near privilege escalation or actions on objectives. The later stages still hold, which is why least privilege, segmentation around sensitive systems, data loss prevention, and behavior analytics matter for insider risk. Limiting what any single account can reach caps the damage an insider can do.
Which kill chain stage is hardest to detect?
Weaponization happens entirely on attacker infrastructure, and passive reconnaissance relies on public sources, so neither leaves traces on your network. Among stages that do touch your environment, command and control is the hardest. Modern C2 runs over encrypted HTTPS, DNS, or legitimate cloud services and checks in at irregular intervals. Egress filtering, which restricts where internal hosts can connect in the first place, is often more reliable than trying to spot the traffic.

