DECEPTION CHECK | PRIVACYKEEPER HUNT STARTERS Hunting templates | October 15, 2026 | Nine read-only query sections START WITH THE QUESTION, THEN CHECK THE DATA Run one numbered section at a time in Microsoft Defender XDR advanced hunting. Sections 1-4 use a 30-day lookback. Sections 5-9 start at seven days. Begin with the affected device and a narrow time window; expand only as needed. Add a DeviceId filter after each table's timestamp filter to scope an investigation. Keep timestamps in a consistent time zone when building the timeline. These tables require the relevant Defender for Endpoint telemetry, permissions, and retention. Check your portal's built-in schema for supported ActionType values. A query error means the query or data source needs attention. Zero hits does not establish a clean device. Positive preflight counts establish available events, not complete collection across every device or every day. WHAT EACH SECTION ANSWERS 1. DeviceFileEvents: exact specimen hashes, package-name leads, and download origin. A file in a download folder, quarantine, or backup is not execution. 2. DeviceProcessEvents / ProcessCreated: launcher execution. Name-only results need validation; process identity includes device, PID, and creation time. Unlike the full report's H2, this starter omits the weak 9P45NL1FPTL8 tag lead: the report does not establish that it reaches a process command line. 3. DeviceImageLoadEvents: the reported modified runtime loaded. This does not establish that its browser-control interface was called. 4. DeviceNetworkEvents: the configured application host appeared in network telemetry. Read ActionType before treating an attempt as successful contact. 5. DNS preflight: available DnsQueryResponse and DnsConnectionInspected records, observation dates, and sample JSON shape. Run an available branch separately if your tenant lacks the other table. Blank ExampleKeys can mean an array or another payload shape, not necessarily an empty event. 6. DeviceEvents / DnsQueryResponse: candidate records containing one of the domains. No undocumented JSON property is assumed. Inspect AdditionalFields to identify the actual query-name field, then build an exact parsed match. Domain text in another field remains a candidate, not a verified lookup. 7. DeviceNetworkEvents / DnsConnectionInspected: exact, case-normalized matches on the documented AdditionalFields.query property, allowing a terminal dot. A DNS lookup is not proof of download, installation, or execution. RemoteIP can identify the resolver. InitiatingProcess fields may be unavailable on these network-inspected events; blank fields mean missing attribution, not that no process ran. Correlate with separate endpoint/network records before attributing the lookup to a process. Encrypted DNS, caching, and collection differences can leave gaps; neither DNS route is a universal inventory of lookups. 8. DeviceRegistryEvents: Chrome/Edge search, home/startup, and extension-policy activity. Whole-key deletions remain visible even with a blank value name. This covers the browser policy roots and their descendants, not deletion of an ancestor such as the entire Software\Policies key. 9. DeviceFileEvents: Chrome/Edge Preferences, Secure Preferences, Web Data and SQLite sidecars, plus extension directories. This catches file-level leads a policy-only hunt misses. It does not show a semantic settings difference. CHECK THE BROWSER CHANGE, NOT JUST THE HIT Keep every writer in sections 8 and 9. UI automation may cause Chrome or Edge itself to write a setting; filtering to PKeep.exe or its children would miss it. Policy refreshes, browser updates, extension updates, sync, and ordinary user choices can produce these events too. A Bing or Google search URL is not an IOC. For a useful before/after comparison, retain a dated baseline for the same user and browser profile. Use approved endpoint or forensic tooling to preserve policy exports, Preferences/Secure Preferences, Web Data, and extension details. Capture SQLite databases consistently, including relevant WAL/SHM files. Analyze copies read-only; do not edit live preferences, launch the suspect package, or visit the listed infrastructure to reproduce the behavior. Compare the effective default search provider and search URL, home/startup targets, extension IDs, install source, and permissions. Check policy source, scope, and mandatory-versus-recommended status in the browser's policy view or a saved export. A recommended policy need not override the user's current choice. DefaultSearchProviderSearchURL also depends on DefaultSearchProviderEnabled; verify the effective policy, not just the presence of a registry value. Compare those findings with approved management changes and user consent. If no earlier baseline exists, record the current state and supporting events without claiming you have proved the previous state or what caused the change. Section 9 recognizes standard Windows Chrome and Edge User Data locations. Custom user-data directories, other browser channels, other operating systems, Firefox, and some atomic replacement/write patterns need separate coverage. Endpoint file telemetry may not record every write. Correlate by device, user, profile, and time; temporal proximity alone is not causation. VALIDATION AND USE The examples have offline matching fixtures, not execution, compilation, or performance validation in a Defender tenant. Check the local schema, test on a limited scope, and review results before adapting them into a detection. Do not automatically block, delete, or remediate from these starter queries. Domains and hashes are matching data, not instructions to browse or download content. PRIMARY REFERENCES Microsoft table schemas: https://learn.microsoft.com/en-us/defender-xdr/advanced-hunting-deviceevents-table https://learn.microsoft.com/en-us/defender-xdr/advanced-hunting-devicenetworkevents-table https://learn.microsoft.com/en-us/defender-xdr/advanced-hunting-deviceregistryevents-table https://learn.microsoft.com/en-us/defender-xdr/advanced-hunting-devicefileevents-table Microsoft's inspected-DNS example documents query, qtype_name, answers, and direction. Its historical Solorigate example uses DnsQueryResponse and a text search of AdditionalFields; it does not document a universal query-name field. https://techcommunity.microsoft.com/blog/microsoftdefenderatpblog/enrich-your-advanced-hunting-experience-using-network-layer-signals-from-zeek/3794693 https://www.microsoft.com/en-us/security/blog/2020/12/18/analyzing-solorigate-the-compromised-dll-file-that-started-a-sophisticated-cyberattack-and-how-microsoft-defender-helps-protect/ Browser policy references: https://learn.microsoft.com/en-us/deployedge/microsoft-edge-policies/defaultsearchprovidersearchurl https://learn.microsoft.com/en-us/deployedge/microsoft-edge-policies/managedsearchengines https://chromeenterprise.google/policies/#DefaultSearchProviderSearchURL https://raw.githubusercontent.com/chromium/chromium/main/components/policy/resources/templates/policy_definitions/DefaultSearchProvider/DefaultSearchProviderSearchURL.yaml https://support.google.com/chrome/a/answer/9844476?hl=en