← All articles
Microsoft Defender XDR · Technical field note

From a suspicious domain alert to process attribution in Defender XDR and Sentinel

Pivot from domain indicators to endpoint network events, initiating processes and evidence-based investigation outcomes.

The alert is the starting point

An August 2025 investigation involved threat-intelligence domain matches associated with an endpoint. The analytical problem was that the presence of a domain in a security alert did not establish whether an endpoint had successfully communicated with an attacker-controlled service or whether a process had executed malicious code.

Here I describe the investigation method using a fictional domain, suspicious-download.example.invalid, and synthetic identifiers. No historical incident outcome is claimed.

Confirm the alert and indicator

In Sentinel, find the relevant SecurityAlert records, their alert provider, time window, entities and extended properties. Inspect the current threat intelligence records using the threat intelligence experience available in your environment. Older examples may use ThreatIntelligenceIndicator; validate the active schema before relying on legacy queries.

kusto
let domain = "suspicious-download.example.invalid";
SecurityAlert
| where TimeGenerated > ago(14d)
| where tostring(Entities) contains domain
    or tostring(ExtendedProperties) contains domain
| project TimeGenerated, SystemAlertId, AlertName,
    ProviderName, AlertSeverity, Entities
| order by TimeGenerated desc

Check source credibility, indicator validity and actual entity matching. A benign parent domain, a subdomain, and a lookalike domain should not be treated as equivalent.

Attribute network activity to processes

In Defender XDR Advanced Hunting, query DeviceNetworkEvents using Timestamp. For the same table in a Sentinel workspace, verify that the connector is enabled and confirm which timestamp field is available before adapting the example.

kusto
let domain = "suspicious-download.example.invalid";
DeviceNetworkEvents
| where Timestamp > ago(14d)
| extend CandidateUrl = tostring(RemoteUrl)
| extend Host = tolower(tostring(parse_url(
    iff(CandidateUrl contains "://", CandidateUrl,
        strcat("https://", CandidateUrl))).Host))
| where Host == domain or Host endswith strcat(".",domain)
| project Timestamp, DeviceId, DeviceName, ActionType,
    RemoteUrl, RemoteIP, RemotePort, InitiatingProcessFileName,
    InitiatingProcessCommandLine, InitiatingProcessSHA1
| order by Timestamp asc

Inspect ActionType to understand the observed event. A DNS observation alone is not proof of a successful connection, and a network connection alone is not proof of a successful HTTP download. A missing hostname may require a documented DNS-to-IP correlation with time bounds; shared hosting and address reuse make raw IP matching noisy.

Recover context around the initiating process

Pivot to DeviceProcessEvents and DeviceFileEvents using the verified device ID, event time and process identity or file hash. Avoid associating all device activity with one alert indiscriminately. For browser activity, review URL and download telemetry where available. DNS resolver or firewall records may provide additional evidence if endpoint DNS events are absent.

kusto
let targetDevice = "synthetic-device-id";
DeviceProcessEvents
| where Timestamp > ago(1d)
| where DeviceId == targetDevice
| project Timestamp, FileName, ProcessCommandLine,
    SHA1, InitiatingProcessFileName, InitiatingProcessCommandLine
| order by Timestamp asc

Adjust the time window to the actual alert, preferably using a narrow interval around the observed event. Consider browser child processes, scripts, installers and new persistence activity. Review any Defender alert and remediation status separately.

Evidence-based disposition

Separate these conclusions: indicator match, DNS resolution, network attempt, successful connection, file creation, execution, and confirmed compromise. Each requires its own evidence. Validate suspicious artifacts with approved internal tools or sandbox workflows without uploading confidential files to public services.

If containment is warranted, follow authorised procedures to isolate affected devices, block relevant indicators and preserve evidence. Check whether the same indicator exists on other endpoints and whether alternate destinations or persistence mechanisms are present.

Validation

In a lab, use an authorised test domain and known process to generate DNS and network activity, then contrast blocked and allowed actions. Confirm what is and isn't visible in Defender, Sentinel and any DNS security platform. Repeat for relevant operating systems.

References