Threat hunting to uncover what scanners miss

Security scanners are excellent at finding what they are built to recognise: exposed services, missing patches, misconfigured controls and malware families linked to current signatures. Their value is speed and repeatability. Yet a clean scanner result cannot prove an environment is uncompromised. Attackers increasingly work through valid remote administration tools, stolen credentials and normal cloud workflows, leaving few artefacts for a point-in-time assessment to flag.

Using threat hunting to find indicators of compromise (IOCs) missed by scanners shifts attention from isolated alerts to sequences of behaviour. A hunter forms a defensible hypothesis, gathers evidence across identity, endpoint, network and cloud systems, then tests whether the activity fits. This matters for Australian enterprises that support critical services and operate across distributed environments, where fragmented telemetry can conceal a patient intruder.

Common blind spots include business-logic abuse, novel malware, fileless activity and persistence hidden in trusted software. A scanner may report no critical vulnerabilities while an attacker quietly exports data through an authorised cloud application. Another attacker can bypass a malicious-file control by living on the land, then use a legitimate VPN to move between systems. These cases demand analytical investigation rather than another signature update.

CARM Security, developed by Exclusive Networks, is designed for post-breach attack remediation and mitigation. Its ecosystem combines technologies from multiple security vendors so enterprises can identify, contain, respond to and remediate cyberattacks with greater coordination. Threat hunting fits naturally around that lifecycle: it uncovers the hidden foothold, supplies evidence for scoping the incident and exposes control failures that need correction after recovery.

Why scanners miss signs of compromise

A scanner evaluates assets against known conditions at a specific moment. Vulnerability scanners look for unsupported software and misconfigurations; endpoint products match files and behaviours against detection logic; threat-intelligence feeds list infrastructure already associated with actors. These methods remain essential because they scale across large estates and create consistent, actionable findings.

An adversary adapts to the controls in front of them. They may rename a tool, remove a file after execution, split activity across accounts or imitate the timing of ordinary administration. A signature-based scanner may still see every individual event as acceptable. The missing indicator is often the relationship among events, such as a new privileged session beginning minutes after an unusual identity change.

There is also a difference between a vulnerability and a compromise. An unpatched internet-facing appliance can be a serious risk without having been used, while a fully patched server can remain compromised through a valid account. Security teams need both vulnerability management and behavioural detection. Treating either as a substitute creates blind spots that become more expensive as dwell time increases.

Building a hypothesis-led hunt

A productive hunt begins with a hypothesis, not a predetermined attacker. Examples include “a compromised user account is creating cloud access from an unfamiliar location”, “an initial-access malware process is establishing persistence outside the normal image” or “mailbox rules are forwarding internal messages to an external destination”. The hypothesis defines the data, time period and technical questions the team must answer.

An investigator then pivots. For a suspected account takeover, the first alert might show an impossible travel event. The next step could compare MFA methods, device fingerprints, OAuth grants, mailbox access and subsequent downloads. Each answer produces a new question, gradually building an attack path. This process is more reliable than collecting unrelated alerts because every data request has a reason and the final finding can be tested.

The hunt needs explicit success criteria and a stopping point. The team should define what evidence would support or disprove the hypothesis, which systems are in scope and how confidence will be recorded. A failed hypothesis can still be valuable: it narrows the likely path, identifies a telemetry gap or shows that an apparently suspicious action was authorised.

Joining telemetry across the enterprise

Threat hunting depends on visibility that crosses ownership boundaries. Identity records reveal authentication and privilege changes; EDR shows process, file and registry activity; DNS, proxy and firewall logs expose destinations; email records show message actions; cloud audit logs capture configuration changes and data access. No single source provides a complete history.

Australian operating environments add practical complexity. Sydney and Melbourne organisations commonly combine Microsoft 365 and other cloud services with on-premises systems, while Brisbane operations may connect mining, logistics or field infrastructure to corporate networks. A hunt confined to the corporate data centre can therefore miss activity in a SaaS tenant or a remote operational network. Endpoint and identity telemetry must be joined according to user, device, address and time.

Sensitive Australian organisations must also preserve evidence quickly. The Australian Customs Service and Australian Border Force illustrate why customs and border environments require careful attention to identity, partner access and connected operational systems. APRA-regulated institutions face notification duties following significant cyber security incidents, making precise scoping important. ASD and the ACSC provide national strategic context, but local defenders still need to interpret it within Australian systems, contracts and managed-service arrangements.

Tracing behaviour that tools overlook

Start with the attack path most consistent with the hypothesis, then work outwards from the earliest confirmed anomaly. Analytical techniques include temporal correlation, identity-sequence analysis, process lineage, rare-destination analysis and retrospective searches across historical logs. The aim is to identify both the entry mechanism and any indicators that a scanner could not match.

Practical hunt priorities

Behavioural pivots

Evidence worth preserving

These outputs become actionable indicators when they are specific enough to validate. They might include a newly identified malicious domain, an attacker-controlled certificate, an unusual OAuth application or a persistence artefact tied to a host. A useful finding explains what the artefact did, which systems it touched and why existing controls did not detect it.

Separating signal from false positives

Threat hunting can generate more noise than it removes if every anomaly becomes an incident. Administrative work, software deployments and emergency access can resemble attacker behaviour. Analysts need to consult change records, service accounts, business owners and known maintenance windows. A false positive is not merely a nuisance; repeated misclassification can train teams to ignore valuable context.

Attackers may also tailor their tradecraft to known detections. A hunt based only on a threat feed can miss renamed infrastructure or a new persistence path. Conversely, a hypothesis based on reliable behaviour can survive changes in tooling. Strong hunts combine intelligence about the actor with evidence generated locally, and they test alternative explanations before declaring a compromise.

Validation should preserve an evidence trail and distinguish initial access from downstream activity. Analysts record where each observation came from, how timestamps were normalised and which links were inferred. They also identify gaps such as missing VPN logs or disabled cloud audit trails. That record supports scoping, helps legal and regulatory teams understand the incident, and shows what must be fixed to support the next investigation.

Connecting discoveries to response

CARM Security gives defenders a coordinated way to bring hunt findings into the wider incident-response process. The integrated vendor ecosystem can help link identity, endpoint, threat-intelligence and containment capabilities, while shared workflows support decisions about affected assets, accounts and data. For Australian buyers, this is relevant in a market where internal teams often depend on a mix of local managed-service providers, global platforms and specialist resellers.

Detection approach Primary question Strongest coverage Typical limitation
Automated scanning Is a known weakness present? Vulnerabilities, configuration errors and known malicious artefacts Point-in-time view and limited visibility of novel tradecraft
Threat intelligence Does an indicator match a known campaign? Attacker infrastructure and documented patterns Campaign changes and unlisted private tooling
Threat hunting Does this behaviour fit a plausible intrusion? Missed indicators, attack sequences and control failure Requires skilled analysts, suitable telemetry and time
Continuous monitoring Is a suspicious event occurring now? Near-real-time alerts and operational context Alerts still require triage and may not reconstruct old activity

A hunt finding should trigger proportionate action. Response teams may isolate a host, revoke sessions, disable a malicious OAuth grant, block confirmed infrastructure and examine related accounts. They must also close the original entry route, remove persistence and determine whether data left the environment. Scanning should resume around repaired systems, while targeted hunts check for missed lateral movement. The objective is verified recovery rather than a false sense of closure.

The priorities worth remembering

The division of labour between tools and analysts is straightforward. Scanners and automated controls provide broad, repeatable coverage; threat hunting supplies the context needed to reveal stealthy activity, validate hidden indicators and explain how an attack progressed. Both are required when the adversary can operate through trusted access and normal business processes.

The point to remember is simple: a clean scanner result means no known weakness was found, not that no attacker is present. Threat hunting closes that gap by connecting the evidence, testing plausible attack paths and turning overlooked compromise into a bounded, actionable incident.