Attack Surface: What It Is, Its Parts and How to Reduce It
Skip to content

Attack Surface

An attack surface is the set of points where an attacker can try to enter a system, affect it or take data from it.

On this page

For an organization, it is the sum of those points across its networks, identities, cloud services, suppliers and people. It changes each time a system, an account or a connection is added, so a description of it holds for the time it was drawn up.

This guide covers the definition, the difference between an attack surface and an attack vector, the parts of an attack surface, how it grows, who owns it, how to measure and reduce it, and where reduction appears in standards.

Attack Surface Defined

The attack surface is everything an attacker can reach and act on from a given starting position. Standards bodies word the definition differently and agree on the idea.

  • NIST: The glossary entry for attack surface, taken from SP 800-53 Rev. 5, defines it as the set of points on the boundary of a system, a system component or an environment where an attacker can try to enter, cause an effect on, or extract data from it.
  • ASD: The cyber security terminology of the Information Security Manual describes the reachable points through which a system could be accessed, manipulated or compromised, including hardware and software interfaces, network services and application programming interfaces.
  • CISA: The NICCS glossary counts digital, physical and human entry points, from websites, servers and APIs to employee devices, shadow IT and social engineering.
  • Microsoft: The endpoint security documentation describes an organization’s attack surface as all the places where an attacker could gain access.

In each definition the surface is made of reachable points, so a system that nothing can connect to adds little to it. The surface is usually divided by kind.

  • Digital attack surface: Hosts, applications, accounts, cloud resources and code can be reached over a network.
  • Physical attack surface: Buildings, network ports and devices can be reached in person.
  • Human attack surface: Staff and contractors can be contacted and deceived.

Attack Surface and Attack Vector

An attack surface is where an attacker could act, and an attack vector is the route or method used against one of those points. A sign-in page is part of the surface. A stolen password entered on that page is a vector.

The Common Vulnerability Scoring System (CVSS) uses the term in this sense. Its Attack Vector metric records the context from which a weakness can be exploited, with the values Network, Adjacent, Local and Physical. A weakness exploitable over the network is open to the widest set of attackers, up to the entire internet.

MITRE ATT&CK groups the entry vectors that attackers use under its Initial Access tactic. Each depends on a particular part of the surface.

Attack vector Part of the surface it uses What reduces it
Exploit of a public-facing application Internet-facing applications and services. Fewer exposed services and prompt patching.
External remote services VPN gateways, remote desktop and other remote access. A small number of managed access paths with multi-factor authentication.
Valid accounts Every account that can sign in, including service and supplier accounts. Removal of unused accounts and multi-factor authentication.
Phishing People, and the mail and messaging channels that reach them. Mail filtering, phishing-resistant sign-in and a habit of reporting.
Trusted relationship Suppliers and partners with access to the environment. Scoped and monitored third-party access.
Supply chain compromise Software, updates and hardware received from others. Fewer dependencies and controlled update sources.

Vectors also combine. NIST describes an attack vector as a path or means by which an adversary can gain access to a system, and SP 800-53 refers to multi-hop attack vectors, where weaknesses on several systems are used in sequence. The internal and identity parts of the surface therefore matter even where the external part is small.

Blocking one vector leaves the surface as it was, while reducing the surface removes a target that many vectors could have used.

Parts of an Attack Surface

An organization’s attack surface divides into parts that different teams own and measure. An intrusion typically enters through one part and reaches data through another.

External

The external surface is everything reachable from the internet, and it is the part an attacker can map without any access. The guide to external attack surface management explains how these assets are found from the outside.

  • Names: Domains, subdomains and DNS records point to systems the organization runs or once ran.
  • Services: Open ports, web applications, APIs and mail servers answer anyone who connects.
  • Access points: VPN gateways, remote desktop and management interfaces offer a way inside.

Internal

The internal surface is what an intruder can reach after a first foothold: file shares, administrative protocols between hosts, unpatched internal systems and local administrator rights. It decides how far one compromised device or account can carry an attacker.

Identity

The identity surface is every account, credential and session that can authenticate: user accounts, administrator roles, service accounts, application registrations, API keys and tokens. A valid sign-in passes through the network controls built for the other parts, which is why unused and over-privileged accounts count as exposure.

Cloud and SaaS

The cloud surface is made of configuration. A resource can be published to the internet by one setting, with no firewall change involved.

  • Public resources: Storage with public access and resources with public addresses are reachable by anyone.
  • Access policies: Permissive roles and long-lived keys widen what one credential can do.
  • Connected applications: Third-party applications granted consent can read mail, files or directory data.

Third Parties

The third-party surface is the access and data held by suppliers: managed service providers with remote access, vendors with VPN credentials, integrations between systems, and software components built into applications. The organization inherits part of each supplier’s surface.

Operational Technology

The operational technology surface is the industrial equipment that can be reached from outside the plant network: programmable logic controllers, human-machine interfaces and remote terminal units, together with the cellular modems and vendor connections used to maintain them. CISA’s guidance on internet exposure asks that necessary remote access passes through a secure gateway, firewall, VPN or centrally managed access solution, with no direct connection to those devices.

People

The human surface is the staff, contractors and executives who can be reached by email, phone, messaging and social media, together with what is public about them. Help desk procedures for password and multi-factor resets belong here, since a caller can invoke them. Domains that imitate the organization extend this part beyond its own systems, as the guide to lookalike domain monitoring explains.

Attack Surface of an Application

A single application has an attack surface of its own, made of every route by which data or commands can enter it. NIST SP 800-53 uses the term at this level, describing attack surfaces as the accessible areas where weaknesses in hardware, software and firmware give an adversary an opportunity.

  • Entry points: Pages, forms, API endpoints, file uploads and message queues accept input from outside the application.
  • Sign-in boundary: Functions offered before sign-in are open to anyone, and functions behind it are open to every account holder.
  • Privileged functions: Administration pages, debug endpoints and support tools carry more authority than the rest of the application.
  • Dependencies: Libraries, frameworks and connected services bring their own code and their own weaknesses.

The control catalog turns this into requirements on the developer. Control SA-11(6) requires attack surface reviews, in which design and implementation changes are analyzed and the attack vectors they create are mitigated. Control SA-15(5) requires the developer to reduce the attack surface to a threshold the organization sets, through measures such as deprecating unsafe functions, reducing the amount of code that executes and removing APIs that are vulnerable to attack.

Each new endpoint, parameter or integration is listed at design time, and a function that ships without a user is removed. Testing of the finished application is covered under web application penetration testing.

How an Attack Surface Grows

An attack surface grows through ordinary work, mostly by additions that are never reversed.

  • New services: A project publishes an application, a test environment or an API, and the test copy stays online after launch.
  • Temporary exceptions: A firewall rule or a remote access path opened for a supplier or a migration remains once the work is done.
  • Accounts that outlive their purpose: Staff leave and projects end, while the accounts, keys and application permissions created for them stay active.
  • Default features: Software ships with services, sample pages and management interfaces switched on, and NIST notes that default functions may be unnecessary for the organization’s operations.
  • Self-service cloud and SaaS: Teams create cloud resources and sign up for applications outside central IT, which is often called shadow IT.
  • Acquisitions: A merger brings in domains, networks and suppliers that the acquiring team has yet to review.

Additions usually have a requester and a deadline. Removals rarely have either, so the surface drifts upward unless removal is made part of the process.

Differences By Size and Role

The parts are the same in every organization, and their weight changes with size.

  • Small organizations: A business with few servers of its own typically holds most of its surface in identity, email and SaaS, along with the remote access its IT provider uses.
  • Mid-sized organizations: A hybrid estate adds remote access gateways, older on-premises servers and a growing set of cloud subscriptions, often with one small team responsible for all of it.
  • Large organizations: Subsidiaries, acquired businesses and separate development teams publish systems independently, so ownership is usually the main gap.

Responsibility divides by role as well.

  • Security lead: The security lead sets the order of work, keeps the counts and reports the trend to management.
  • IT manager: The IT manager makes the changes, from firewall rules and DNS records to disabled accounts.
  • System and application owners: Each owner confirms whether an exposed system is still needed and accepts the risk of what stays.

Attack Surface Shared With an MSP

An organization that uses a managed service provider (MSP) shares part of its attack surface with that provider, and the provider’s accounts and tools are a route into every client it serves. A joint advisory on threats to MSPs and their customers, issued in May 2022 by CISA, the NSA, the FBI and their counterparts in Australia, Canada, New Zealand and the United Kingdom, sets out measures for both sides.

For the customer, the advisory concentrates on the provider’s accounts.

  • Scoped accounts: Provider accounts reach only the systems the provider manages and stay out of internal administrator groups.
  • A dedicated path: Provider traffic arrives through a dedicated VPN or another secure access method and is limited to that connection.
  • Contract end: Provider accounts are disabled once the provider stops managing the infrastructure, a step the advisory notes can be overlooked when a contract terminates.

For the provider, the measures keep one client’s compromise from becoming another’s.

  • Separation between clients: Customer data sets are segregated from each other and from the provider’s internal network, and administrator credentials are unique to each customer.
  • Least privilege: Provider staff hold only the rights a task needs, and every account with access to a customer environment uses multi-factor authentication.

Measuring an Attack Surface

An attack surface is measured by counting reachable points from records, part by part, and comparing the counts over time. A short set of counts taken on a schedule is typically more useful than a single score.

Part Record to pull Where it comes from What to count
External Public DNS records and certificates issued for the organization’s domains. A zone export from the DNS provider, and Certificate Transparency logs searched by domain. Hostnames that resolve, and hostnames absent from the asset inventory.
External Open ports and services on each public address. An external scan with a tool such as nmap, and the inbound rules exported from firewalls and load balancers. Services reachable from the internet, and the management interfaces among them.
Cloud Resources with public addresses or public access. The cloud provider’s inventory query service, for example Azure Resource Graph across subscriptions. Public endpoints, and storage that allows anonymous access.
Identity Last sign-in for each account. The user list in the Microsoft Entra admin center with the last interactive sign-in time column added, or the equivalent in another directory. Enabled accounts with no sign-in inside the review window, which Microsoft’s guidance on inactive users puts at 90 to 180 days in many organizations.
Identity Multi-factor registration. The Activity report under Authentication methods in Microsoft Entra, which shows whether each user is MFA capable. Accounts able to sign in with a password alone.
SaaS Applications granted access to data. The enterprise applications list and the consent records in the identity provider. Applications with mail, file or directory permissions, and who approved each one.
Internal Listening services and installed software on each host. Authenticated vulnerability scanning and the endpoint management inventory. Hosts running services outside the approved baseline.
Third parties Suppliers with access. The supplier register and the VPN and remote access account lists. Supplier accounts, and those without a named internal owner.

The interval between counts matters as well. External counts go stale fastest, and the external attack surface management guide lists the published minimums for how often discovery should run.

Beside the counts, a few trend measures show whether the surface is shrinking.

  • Net change: The internet-reachable services added in a period are set against those removed.
  • Unowned share: The proportion of exposed assets without a named owner is tracked toward zero.
  • Expired exceptions: Firewall exceptions, temporary accounts and test environments past their end date are listed with their owners.

Attack Surface Reduction in Order

Attack surface reduction means removing reachable points first and protecting the ones that must remain second. NIST describes it in SP 800-53 as reducing risk by giving attackers less opportunity to exploit weaknesses, through measures that include least privilege, least functionality, fewer entry points available to unauthorized users and less code that executes.

The order below starts with internet exposure and works inward. CISA’s Internet Exposure Reduction Guidance follows the same sequence: assess what is exposed, decide whether each exposure is needed, protect what must remain, and repeat the assessment.

  1. Establish what is exposed. Build the list of internet-reachable assets from the records above and give each one an owner. Include the remote access that integrators, managed service providers and vendors hold.
  2. Remove what has no owner or purpose. Decommission forgotten hosts, test environments and unused DNS records.
  3. Take management interfaces off the internet. Administration interfaces for firewalls, routers, VPN devices and servers should be reachable only from an internal network or through a separate access control point. CISA’s Binding Operational Directive 23-02 gives federal agencies 14 days to do this once such an interface is found.
  4. Consolidate remote access. Route remote access through a small number of managed entry points with multi-factor authentication. NIST notes that limiting the number of access control points for remote access reduces attack surfaces.
  5. Disable unused accounts and reduce privilege. Disable expired and inactive accounts, remove standing administrator rights and review application permissions.
  6. Remove unneeded functions. Uninstall unused software and close unused ports and protocols on the systems that remain. This is least functionality, control CM-7 in SP 800-53, under which a system provides only the capabilities the organization needs.
  7. Harden what must stay exposed. Change default passwords, apply security patches, replace software and devices that no longer receive security support, and monitor inbound and outbound traffic.
  8. Segment internally. Separate networks so that one compromised device reaches fewer systems.

Removal comes before protection because a removed service needs no patching, monitoring or testing afterward.

ASD takes the same line in the ASD Essentials for enterprise IT, where approved configuration baselines disable or remove unnecessary services, insecure settings and unsupported or weak communication protocols.

Deciding What Stays Exposed

Each exposed system gets a recorded decision: remove it, restrict it, or keep it with protection. CISA’s guidance makes operational need the test, so an asset stays internet-accessible only where operations require it.

The record states what the system does, who has to reach it and whether the product still receives security updates. The outcomes below run from the most reduction to the least, and the first one the business can accept is chosen.

  1. Remove. A system with no current purpose is shut down, and its DNS record, firewall rule and certificate are retired with it.
  2. Move behind an access point. A system used only by staff or a supplier is placed behind the remote access gateway or a jump host, so it stops answering the internet directly.
  3. Restrict by source. A system used by a known set of partners is limited to their addresses.
  4. Keep and harden. A system that serves the public stays reachable, with patching, multi-factor authentication where it applies, and monitoring.

Some systems can be neither patched nor removed. CISA’s performance goals call for compensating controls such as segmentation or a firewall in that case, and NIST notes that keeping unsupported components off public networks reduces their risk. The systems that stay exposed form the natural scope for penetration testing.

A First Reduction Cycle

A first cycle is a short project with a count at each end. The sequence below follows a mid-sized professional services firm with one office network, a cloud tenant for mail and files, and an outside IT provider.

  1. Week 1, count. The IT manager exports the DNS zone, searches Certificate Transparency logs for the firm’s domain and scans its public addresses. Beside the website, the mail gateway and the VPN, the list shows a file transfer server from a finished project, a staging copy of the client portal and the firewall’s administration page.
  2. Week 2, assign owners. A named owner is sought for each item, and none is found for the staging site or the file transfer server.
  3. Week 3, remove and restrict. Both unowned systems are shut down and their DNS records deleted. The firewall administration page is limited to the internal network, and the IT provider moves to named accounts that connect through the VPN with multi-factor authentication.
  4. Week 4, review identity. Enabled accounts with no sign-in inside the review window, among them former staff and a service account for a retired scanner, are disabled. Applications unused for more than 90 days, which Microsoft Entra lists under its recommendation to remove unused applications, are reviewed with their owners.
  5. Week 5, harden and record. The systems that stay exposed are patched and their logs are confirmed to reach monitoring. The counts are taken again and stored beside the first set, with an owner and a reason against each remaining exposure.

The second set of counts becomes the baseline for the routine checks that follow.

Attack Surface Reduction on Endpoints

On Windows devices the phrase also names a specific Microsoft feature. Attack surface reduction rules in Microsoft Defender Antivirus target software behavior that attackers commonly abuse through malware.

  • Downloads and launches: Executable files and scripts attempt to download or run other files.
  • Obfuscated scripts: A script is obfuscated or otherwise untrusted.
  • Child processes: An application that is often exploited, such as an Office app, creates another process.
  • Code injection: One process injects code into another.

Legitimate applications sometimes behave the same way, so a rule can run in audit mode, which records what would have been blocked. Microsoft recommends block mode for its standard protection rules and testing in audit mode for the others before they are enforced.

Reduction in Standards and Frameworks

Attack surface reduction appears in control catalogs as specific requirements, so a reduction program can be mapped to obligations an organization may already have. The references below come from NIST SP 800-53 Rev. 5 and CISA’s Cross-Sector Cybersecurity Performance Goals 2.0.

Reference What it asks Record that shows it
SP 800-53, CM-7(1) Systems are reviewed at a defined frequency for unnecessary or nonsecure functions, ports, protocols, software and services, which are then disabled or removed. A dated review listing what was removed.
SP 800-53, AC-2(3) Accounts are disabled within a defined period once they have expired, are no longer associated with a user or have been inactive. Disabled accounts set against the leaver list.
CPG 3.S, Secure Internet Facing Devices Internet-facing assets are minimized, and network management interfaces are kept off the public internet. External scan results compared over time.
CPG 3.D, Revoking Credentials for Departing Staff A defined offboarding process covers staff, contractors and vendors, and accounts are disabled after a specified period of inactivity, with 30 days given as an example. Offboarding tickets matched to account status.
CPG 1.E, Manage Risks from Managed Service Providers The risks posed by a managed service provider are identified, recorded, assessed, prioritized and monitored. The supplier register entry for each provider.

CISA describes the CPGs as voluntary practices, and SP 800-53 is the control catalog for US federal systems. Expectations on testing frequency are set out under continuous penetration testing.

Keeping the Surface Reduced

A reduced attack surface stays reduced where change is watched and removal is part of routine work. The events that grow the surface are the points at which to check it.

Event Check
A new hostname or public address appears Confirm the owner, the purpose and the services exposed.
A project or migration closes Remove its test environments, firewall exceptions and temporary accounts.
A person or supplier leaves Disable the accounts, revoke the keys and remove the remote access.
A new advisory covers an exposed product Confirm which assets run the product and whether they are reachable.
A scheduled review falls due Compare the current counts with the last review and explain each increase.

Several practices keep those checks from depending on memory.

  • Expiry dates: Every exception, temporary account and test environment is created with an end date and an owner.
  • Scanning and testing: Scans find known weaknesses on the assets that remain, and tests confirm which of them can be used in practice.
  • A program: Continuous threat exposure management joins discovery, prioritization and validation into a repeating cycle.

Reduce and Test With ThreatDefence

Reduction leaves a set of systems that have to stay reachable, and ThreatDefence tests that remainder as it changes. The testing belongs to the same platform ThreatDefence uses to automate security monitoring, compliance and cyber risk management, an AI SOC with analysts on duty 24x7.

  • What is exposed: In the attack surface management product, the ASM agent maps the domains, hosts, services and certificates that face the internet, which shows what there is to remove.
  • What changed: Every change is recorded, and a host that appears or changes can trigger a test.
  • What can be used: Through agentic penetration testing, the testing agent confirms which weaknesses on the remaining systems are exploitable and keeps the proof.
  • What was fixed: Each problem has one tracked action item with fix steps, and the agent retests once a fix is reported.

The Brand agent watches for lookalike domains and leaked credentials. Internal network testing and social engineering testing sit outside this service, and an independent human-led engagement can cover them.

Frequently Asked Questions

What is an attack surface in cybersecurity?

In cybersecurity, the attack surface is everything about a system or an organization that an attacker can reach and try to use. NIST places it at the boundary of a system, a component or an environment. For an organization that means internet-facing services, accounts, cloud configuration, supplier access and the people who can be contacted.

What is the difference between an attack surface and an attack vector?

The surface is the target and the vector is the technique. An exposed remote desktop service belongs to the attack surface, and password guessing against it is an attack vector. Closing the service removes the target for every technique that depended on it, which is why surface reduction usually comes before work on individual vectors.

What is a digital attack surface?

A digital attack surface is the part of an attack surface made of technology: domains, hosts, applications, APIs, accounts, cloud resources and code. It is usually set beside the physical surface, such as buildings and ports, and the human surface. Most measurement work concentrates on the digital part, since its records can be pulled and counted on a schedule.

What is attack surface reduction?

Attack surface reduction is the work of removing reachable points and limiting what the remaining ones allow. The first actions in most environments are decommissioning unused assets, taking management interfaces off the internet and disabling unused accounts. Hardening, patching and segmentation then apply to the systems that have a reason to stay.

Are attack surface reduction rules the same as attack surface reduction?

No. Attack surface reduction rules are a feature of Microsoft Defender Antivirus that blocks risky software behavior on Windows devices, such as Office apps creating child processes. Attack surface reduction as a practice is wider and covers internet exposure, accounts, cloud configuration and supplier access. The rules are one endpoint control within it.

Can an attack surface be eliminated?

No. Any organization that sends email, serves customers online or lets staff sign in remotely keeps some reachable points. The aim is a surface that is known, owned and as small as the business allows, with the remaining points patched, protected by multi-factor authentication and monitored.

How often should an attack surface be reviewed?

A review on change plus a fixed schedule suits most environments. The external part is usually checked most often, because a new hostname or open port can appear on any day. Account, application permission and supplier access reviews often run monthly or quarterly, depending on how quickly staff and suppliers change.

What is the difference between an attack surface and a vulnerability?

An attack surface is a set of reachable points, and a vulnerability is a weakness in one of them. A patched, well-configured VPN gateway is still part of the surface, because a future flaw or a stolen credential could be used against it. Reducing the surface lowers the number of places where a vulnerability would matter.

What is attack surface analysis?

Attack surface analysis is the review of every point where a system accepts input or grants access, carried out to decide which points can be removed or restricted. For an application it lists entry points, privileged functions and dependencies, and NIST SP 800-53 asks developers to review them when the design or implementation changes. For an organization it is the count of reachable points in each part of the surface.

Does moving to the cloud reduce the attack surface?

No, in most cases a move to the cloud changes the surface more than it shrinks it. Servers and network equipment that the organization used to patch may go, while identities, access policies, public resource settings and connected applications take their place. The cloud part is made of configuration, so it is measured from the provider’s inventory and the identity directory.

Talk to Our Experts

See how ThreatDefence automates security monitoring, compliance and cyber risk management in one platform.

    I'm an MSP, MSSP, or reseller

    By submitting this form, you accept the terms in our privacy policy.