EASM and CAASM are both sold as "attack surface management", yet one starts from the internet and works inwards while the other starts from your own tools and works outwards, so buying one and expecting the other's answers leaves a predictable blind spot. External attack surface management (EASM) discovers internet facing assets that can be attributed to you, using the same public data an attacker uses: DNS, certificate logs, registration records and internet wide scanning. Cyber asset attack surface management (CAASM) pulls asset records from the systems you already run, through their APIs, and correlates them into one inventory with ownership and control coverage. This guide compares the two as data sources: the question each answers, how each is built, how each fails, how to join them, and which to fund first. It is written for exposure management leads, vulnerability management owners and security architects who keep being asked whether they need one, the other or both.
Key takeaways
- Different starting points, different blind spots: EASM sees what is reachable from the internet whether or not you registered it; CAASM sees what your tools already know about, whether or not it is reachable.
- The gap between them is the finding: an internet facing host that EASM attributes to you but CAASM cannot match to any agent, CMDB record or cloud account is the highest value output of running both.
- Attribution is EASM's weak point, coverage is CAASM's: EASM guesses ownership from public records and gets it wrong in both directions; CAASM is only as complete as the connectors you license and the tools you have deployed.
- Neither tests exploitability in depth: both produce inventory and exposure signals that feed vulnerability scanning, validation and the wider exposure management programme, not replace them.
- Sequence by your biggest unknown: organisations with a sprawling, acquired or marketing driven internet presence usually start with EASM; organisations with many overlapping internal tools and no trusted inventory usually start with CAASM.
The short answer
EASM answers "what of ours can an attacker reach that we did not know about?" CAASM answers "across everything our tools report, what exists, who owns it and which controls are missing?" They overlap only on internet facing assets that are also well registered internally, which is the part of the estate that matters least for discovery. Most organisations past a few hundred staff eventually need both data sets in one inventory, because the NCSC asset management guidance places internet facing presence (domain registrations, IP address ranges, digital certificates and social media accounts) inside asset management scope, alongside the internal estate. The practical question is which data source to build first and how to make each correct the other.
This page goes deeper than the overview in the Attack Surface Management Guide. For how EASM output feeds scanners, see the EASM vs Vulnerability Scanning comparison; for how a CAASM inventory differs from cloud configuration checking, see CAASM vs CSPM. Neither of those topics is repeated here.
Side by side
| Dimension | EASM | CAASM |
|---|---|---|
| Direction | Outside in: starts from seeds (brand names, root domains, ASNs, known IP ranges) and expands | Inside out: starts from authenticated API connectors to tools you own |
| Primary data | Passive DNS, certificate transparency logs, WHOIS and RDAP, routing data, internet wide port and banner scans, web crawling | EDR, MDM, directory, CMDB, cloud provider inventories, vulnerability scanners, SaaS admin APIs, HR and procurement feeds |
| Credentials needed | None for discovery; seeds and confirmation of ownership | Read only API credentials to every connected source |
| Typical finding | Unregistered subdomain, forgotten cloud host, exposed admin interface, expired or mis-issued certificate, lookalike domain | Device with no EDR agent, server missing from CMDB, account with no owner, cloud account outside the landing zone, scanner coverage gap |
| Ownership signal | Inferred from registration, certificate subject, hosting and naming patterns | Read from source systems: tags, CMDB owner fields, directory managers |
| Main error mode | Mis-attribution (false positives and missed assets) and stale scan data | Missing connectors, stale records, poor correlation creating duplicates or merging distinct assets |
| What it cannot see | Anything not reachable or not publicly recorded: internal hosts, endpoints behind NAT, most identities | Anything no connected tool knows about: shadow IT, third party hosted sites, assets in unknown cloud accounts |
| Natural owner | Exposure or vulnerability management, with the web and DNS teams | Asset management or security operations, with IT operations and the CMDB owner |
EASM in depth: threat model, architecture and failure modes
Threat model
EASM models the reconnaissance phase of an external attacker. MITRE ATT&CK describes it as T1595 Active Scanning and T1596 Search Open Technical Databases, including T1596.003 Digital Certificates, followed by initial access through T1190 Exploit Public-Facing Application or T1133 External Remote Services. An organisation that does not know which edge devices it runs cannot patch, isolate or check them in time when a flaw is disclosed. The Ivanti Connect Secure exploitation in 2024 is a clear example: the first defensive question was simply where the internet facing gateways were and what credentials they held.
Architecture in words
An EASM service takes seeds you confirm, then expands recursively. Root domains lead to subdomains through passive DNS and certificate logs; certificates lead to further names in their subject alternative names; IP addresses lead to netblocks and hosting providers; web pages lead to linked hosts and technologies. Each candidate asset receives an attribution confidence. Confirmed assets are then fingerprinted: open ports, service banners, TLS configuration, software versions and screenshots. Certificate Transparency is central here. RFC 9162 (Experimental, December 2021, obsoleting RFC 6962) describes public, append only logs of TLS server certificates and "monitors" that watch logs for certificates issued for their domains, which is exactly the feed EASM tools consume. See the Certificate Transparency Monitoring Guide for Web Teams.
Failure modes
- Over attribution: a shared hosting IP, a similarly named company or a former subsidiary is pulled in, and analysts burn time disproving ownership.
- Under attribution: sites built by agencies on their own accounts, or acquisitions registered under a different legal entity, never connect back to your seeds.
- Snapshot staleness: scan data may be days old, so short lived services are missed and fixed exposures reappear.
- Version guessing: banner based vulnerability matching produces false positives where vendors backport fixes without changing version strings.
- No owner, no action: findings land in a queue with a hostname and nothing else, because EASM has no reliable owner field.
CAASM in depth: threat model, architecture and failure modes
Threat model
CAASM models the defender's own blind spots rather than an attacker's path. The risk it addresses is assets that exist inside the estate but sit outside controls: a laptop without an EDR agent, a server no scanner covers, a cloud account created on a corporate card, a SaaS tenant with local admin accounts. The Midnight Blizzard intrusion at Microsoft began with a password spray against a legacy, non-production test tenant account without MFA, followed by abuse of a legacy test OAuth application: an asset that existed, was reachable and sat outside the controls applied to production. Finding that class of asset is an inventory problem first.
Architecture in words
A CAASM platform connects to source systems with read only API credentials, pulls their asset and identity records on a schedule, normalises fields, and correlates records into a single entity per asset using keys such as hostname, serial number, MAC address, cloud instance ID and user principal name. It then runs queries across the merged view: devices seen by the directory but not by EDR, cloud instances absent from the scanner, accounts with no manager. NCSC guidance describes the same pattern without the product label: combine procurement records, device management, logging, vulnerability scanners, manual entries and certificate records; normalise into one authoritative record; record a "last seen" time or confidence per record; and use differences between sources to reveal inconsistencies. The Asset Inventory and CAASM Buyers Guide covers connector evaluation.
Failure modes
- The connector ceiling: CAASM cannot discover what no connected tool has seen. Shadow IT bought outside procurement, and third party hosted web properties, stay invisible.
- Correlation errors: reused hostnames, cloned virtual machine images and recycled IP addresses either duplicate assets or merge two different ones, so counts and gap reports are wrong.
- Over privileged connectors: the platform holds API credentials to dozens of systems, making it an attractive target itself (see the hardening notes below).
- Gap reports without owners: queries surface thousands of missing agents, but nobody is assigned to close them, so the dashboard becomes a museum.
- Ephemeral and SaaS drift: containers and short lived instances churn faster than sync intervals, and SaaS applications appear only if someone connects them.
Where the two views meet: reconciliation
The value of running both comes from comparing them. Feed EASM into the CAASM inventory as a source and work the resulting sets.
| Set | Meaning | Typical action |
|---|---|---|
| In EASM, not in CAASM | Internet facing asset attributed to you that no internal tool knows: shadow IT, agency sites, forgotten cloud accounts, acquisition estates | Confirm ownership; register or decommission; bring the hosting account under management; add to scanner scope |
| In CAASM, internet facing, not in EASM | Asset your tools mark as public that external discovery did not attribute | Add as an EASM seed; check whether it is genuinely exposed or the internal record is wrong |
| In both | Known, reachable asset | Enrich EASM findings with the CAASM owner and control status so exposure tickets route automatically |
| In CAASM only, internal | Internal estate, out of EASM's view by design | Work control coverage gaps (agent, scanner, backup, logging) through CAASM queries |
Matching keys: join on fully qualified domain name, public IP address plus port, certificate fingerprint and cloud resource ID. Cloud IPs are recycled, so record the time window of each match.
Ownership flows one way: CAASM supplies the owner; EASM supplies the exposure. An exposure ticket without an owner from the inventory should be treated as an inventory defect, not an exposure backlog item.
Billing as a tie breaker: NCSC suggests reconciling procurement records and cloud billing against the inventory to find assets discovery missed, and notes that free cloud services may not appear in procurement records at all. That is exactly where EASM earns its keep.
Where identity fits
Both tools are now marketed with identity features, and the split matters for non-human identities. EASM can surface exposed identity endpoints (login portals, legacy authentication endpoints, exposed API gateways) and leaked credentials in public code, but it cannot tell you which service account or workload identity sits behind them. CAASM can ingest identity provider, cloud IAM and SaaS records, so it can show accounts without owners, but most platforms model identities as attributes of assets rather than as assets with their own lifecycle. Neither replaces a dedicated non-human identity inventory with owners, credentials and rotation status. Use CAASM correlation to link workloads to the identities they run as, and see NHI Discovery Across Cloud, SaaS, CI/CD and On-Premises Environments and NHI Inventory Reconciliation for the identity side.
Secure the CAASM platform itself: issue each connector a dedicated, read only service identity scoped to inventory APIs, store its secrets in a vault, rotate them, alert on use outside the platform's egress addresses, and review connector scopes when vendors add write capable features. A CAASM console with standing read access to EDR, IdP and cloud APIs is a map of your estate for whoever compromises it.
Worked scenario: a professional services firm after two acquisitions
A 4,000 person firm runs an EDR platform, an MDM, a directory, a CMDB nobody trusts, two cloud providers and a quarterly authenticated scan. It acquired two smaller firms in eighteen months. Figures are illustrative.
- Month 1, EASM pilot: seeds are the three brand names, 40 known root domains and the main ASN. Discovery returns about 1,900 candidate hosts; after attribution review, 1,150 are confirmed. Of those, 310 are not in the CMDB or any cloud account the central team manages, mostly agency built microsites, an acquired firm's file sharing service and a test environment on a personal cloud account.
- Month 2, triage: eleven hosts expose management interfaces (SSH, RDP and appliance web consoles). These are closed or put behind the zero trust access broker within 14 days, using the remediation options in CISA BOD 23-02 as a benchmark even though the directive binds only US federal agencies.
- Months 3 to 5, CAASM rollout: connectors to EDR, MDM, directory, both clouds, the scanner and the CMDB. The first gap query finds 9% of directory joined devices with no EDR agent and 140 cloud instances outside scanner scope. EASM output is ingested as a source, which lets the 310 unregistered hosts be tracked to closure with owners.
- Month 6, steady state: EASM runs continuous discovery with weekly attribution review; CAASM syncs daily; new external assets without an internal match open a ticket to the business unit that owns the brand. The unregistered external host count falls below 30.
EASM went first because the acquisitions had brought internet facing assets nobody inside knew about; a firm with one stable web estate and many overlapping tools would have started with CAASM.
Decision criteria
| Situation | Lean towards | Why |
|---|---|---|
| Recent acquisitions, many brands, agency built sites | EASM first | Unknown external assets are the largest risk and no internal tool knows them |
| Many security tools, conflicting asset counts, no trusted CMDB | CAASM first | Correlation of existing data gives fast control coverage wins |
| Small organisation, one web estate, few tools | Neither as a product yet | Certificate transparency monitoring, DNS zone review and a maintained spreadsheet or CMDB may suffice; revisit as tool count grows |
| Regulated environment needing evidence of complete inventory | Both, CAASM as system of record | Auditors ask for coverage evidence across the whole estate, internal and external |
| Formal exposure management programme | Both as discovery inputs | The external surface is a common first scope; internal coverage gaps follow. See CAASM vs CTEM |
Buying inside a platform: some exposure platforms bundle both. Test them separately anyway: run the EASM module against seeds you choose and count confirmed unknown assets, and connect the CAASM module to your real tools and check correlation accuracy on a sample of 100 assets you know well.
Implementation sequence
- Agree the inventory of record: decide whether the CMDB or the CAASM platform holds the authoritative asset record, and which fields each source owns.
- Prepare EASM seeds and confirm them with the business: legal entity names, brands, root domains, ASNs, registrar accounts. Include acquired entities and recently divested ones you need to exclude.
- Pilot EASM for 30 days with manual attribution review: measure false attribution rate before automating ticket creation.
- Connect CAASM to three high quality sources first: typically directory, EDR and the main cloud provider, then add sources one at a time and check correlation after each.
- Ingest EASM into CAASM: treat it as a source and build the reconciliation queries in the table above.
- Route by owner: automate tickets only for assets with an owner; send ownerless assets to an inventory triage queue with a weekly review, in line with CIS Safeguard 1.2's weekly process for unauthorised assets.
- Roll back safely: both tools are read only discovery, so rollback means pausing ticket automation or a noisy connector, not reverting changes. Keep active EASM scanning off sensitive OT ranges, as NCSC notes active methods may be unsafe there.
Detection and measurement
- Unregistered external assets: count of confirmed internet facing assets with no internal match, and median days to registration or decommission.
- Attribution precision: share of EASM candidate assets confirmed as yours after review; a falling rate means seeds are too broad.
- Control coverage: percentage of CAASM assets with EDR, scanner coverage and an owner, by business unit.
- Record freshness: share of assets with a "last seen" within the expected window for their type, as NCSC suggests.
- Exposed management interfaces: count and age of internet reachable admin services (SSH, RDP, appliance consoles).
- Connector health: failed syncs per week and age of the oldest successful sync per source.
Benchmarks for cadence exist in public guidance. CISA BOD 23-01 requires US federal agencies to perform automated asset discovery every 7 days and vulnerability enumeration every 14 days, and excludes ephemeral assets such as containers and third party SaaS from scope, which are precisely the areas where a CAASM sync and SaaS connectors add value. CIS Controls Safeguard 1.3 sets active discovery daily or more often for IG2 and IG3, and Safeguard 1.5 passive discovery at least weekly for IG3. For an outside view at no cost, eligible organisations can use the CISA Cyber Hygiene vulnerability scanning service, which scans public static IPv4 addresses and reports weekly; compare its findings with your EASM output.
Framework mapping
| Framework | Reference | EASM contribution | CAASM contribution |
|---|---|---|---|
| NIST CSF 2.0 | ID.AM-01, ID.AM-02 | Finds hardware, software and services exposed externally that are missing from inventories | Maintains correlated hardware, software and service inventories |
| NIST CSF 2.0 | ID.AM-04 | Surfaces supplier hosted services carrying your brand | Records supplier services through SaaS and procurement connectors |
| NIST CSF 2.0 | ID.RA-01 | Exposure indicators on external assets | Imports scanner findings per asset |
| NIST SP 800-53 Rev. 5 | CM-8, PM-5 | Discovery input to component and system inventory | System component inventory with ownership |
| NIST SP 800-53 Rev. 5 | RA-5, CA-7 | Continuous external monitoring feeding scanning scope | Coverage verification for scanning and monitoring |
| CIS Controls v8.1 | Control 1 (Safeguards 1.1 to 1.5), Control 2 | External discovery and unauthorised asset detection | Detailed enterprise asset and software inventory |
| ISO/IEC 27001:2022 Annex A | A.5.9 Inventory of information and other associated assets | Completeness check from outside | The inventory itself |
| MITRE ATT&CK | T1595, T1596, T1190, T1133 | Mirrors attacker reconnaissance to shrink initial access surface | Indirect: ensures exposed assets have owners and controls |
Common failure patterns
- Buying EASM and calling it the inventory: it covers only what is internet facing and attributable, so internal control gaps stay unmeasured.
- Buying CAASM and assuming it discovers shadow IT: it correlates what tools already see; assets bought and hosted outside IT remain invisible until an external view adds them.
- Automating tickets before measuring attribution: a 20% false attribution rate sent straight to application teams destroys trust in the programme within weeks.
- Forgotten non-production estates: test tenants, demo environments and legacy apps sit outside both seeds and connectors, the pattern behind the Midnight Blizzard initial access. Seed EASM with test and staging domains and connect CAASM to every tenant, not just production.
- Treating edge appliances as network kit, not assets: VPN gateways and firewalls missing from the inventory slowed response to mass exploitation such as the Ivanti Connect Secure campaign.
Practitioner checklist
- Write down the question you need answered first: unknown external assets, or internal control coverage.
- Confirm EASM seeds with legal and brand owners, including acquisitions and test domains.
- Measure EASM attribution precision during a 30 day pilot before automating anything.
- Connect CAASM to directory, EDR and primary cloud first, then add sources one by one and spot check correlation on 100 known assets.
- Ingest EASM findings into the inventory and build the "external but unregistered" query.
- Route exposure tickets only through owners taken from the inventory; review ownerless assets weekly.
- Give each CAASM connector a dedicated read only identity, vault its secret and rotate it.
- Report the metrics above monthly to the risk owner.
Standards and references
- NCSC: Asset management
- CISA: BOD 23-01, Improving Asset Visibility and Vulnerability Detection on Federal Networks
- CISA: BOD 23-02, Mitigating the Risk from Internet-Exposed Management Interfaces
- CIS: Controls Assessment Specification, Control 1 Inventory and Control of Enterprise Assets
- IETF: RFC 9162 Certificate Transparency Version 2.0
- NIST: SP 800-53 Rev. 5, Security and Privacy Controls for Information Systems and Organizations
- CISA: Cyber Hygiene Services
- MITRE ATT&CK: T1596 Search Open Technical Databases
Related NHI Mgmt Group resources: EASM vs Vulnerability Scanning, CAASM vs CSPM, CAASM vs CTEM, Attack Surface Management Buyer's Guide, Asset Inventory and CMDB Security Guide.