Turn unfamiliar rows into answerable questions

Investigate DMARC Failures and Source IPs

An unfamiliar source IP in a DMARC report is a question, not an identification. It may relate to an approved service, a changed sending route, a forwarding system, unauthorized domain use, or something your current records cannot explain. Start with what the receiver actually reported, measure the affected volume, and compare that evidence with an authorized sender inventory. A failure deserves attention without turning an address into a named attacker or treating a recognizable service as automatically approved.

Use the DMARC Report Analyzer to inspect supplied reports locally. Choose Files for report attachments or Paste XML for report text, then select Analyze reports. Load example fills the input only. The results organize declarations from the supplied XML; they do not independently authenticate the report, establish who operated an IP, or determine whether a message reached an inbox.

Keep the receiver, policy domain and reporting window attached

Before investigating a source, identify the report that contains it. Read the reporting organization, contact, report ID, policy domain, and beginning and ending timestamps. Those fields give the row its declared context. The same address in a different receiver report or another window is another observation, not necessarily the same incident. Treat the organization and contact as supplied metadata rather than independently verified authorship. Keep a private investigation note that names the report and UTC window so another authorized reviewer can find the same evidence without guessing from a filename. An attachment name is convenient filing information, not a substitute for the identity inside the report.

Separate the published policy domain from the row’s header_from domain and from domains listed under raw authentication results. They answer different questions: which policy context was reported, which author domain appeared in the grouped messages, and which identities were associated with authentication checks. A shared source address can appear in several such contexts. Do not collapse those domains into one sender label or transfer a conclusion about one domain to another. If your organization operates several brands or subdomains, ask which business team was authorized to use each author domain during the actual window, rather than using today’s account list as historical proof.

Receiver-specific reports can expose different parts of a mail stream. Read the Google DMARC reports guide, Microsoft DMARC reports guide, or Yahoo DMARC reports guide when the collection context matters. A missing report from another receiver is missing coverage, not evidence that nothing happened there. Likewise, a source absent from this window might have sent only outside it or to receivers not represented in your files. Describe the finding as present or absent in the supplied reports, preserving the narrower claim instead of turning a partial sample into a complete organizational mail inventory.

Prioritize reported volume instead of counting visible rows

A report row represents the number in count, not one individual email. Suppose an invented source has a failed row with count 900 and a passing row with count 100. That is 1,000 reported message instances, of which 900 are associated with the failing row. It is not a fifty percent failure rate merely because one of two rows fails. Conversely, a source with many tiny failure rows may account for fewer instances than one large failure row elsewhere. Use count-weighted totals to decide where an administrator’s first investigation is likely to explain the most reported volume, while still retaining low-volume anomalies that have a concrete business consequence.

The derived DMARC summary follows the reported aligned method results: either aligned method passing gives pass; both failing gives fail; any remaining combination is unknown. Missing or unsupported aligned values remain unknown with a warning. Do not move them into failure to make the figures look decisive, or into pass because a raw authentication entry looks reassuring. Record the unknown count beside the known outcomes. A zero-count row contributes no message instances even though the row itself exists. When a denominator is zero, a percentage is unavailable rather than zero percent; there is no measured population from which to calculate that rate.

Keep the population behind every percentage visible in your notes. If you select one report or inspect one source, its counts describe that subset, not the entire batch or all mail sent by the domain. A total of reported instances is not a count of unique emails, successful deliveries, recipients, or messages your organization originated. Receivers may observe the same underlying message along different paths, and aggregate rows do not contain a universal message identifier for resolving that. The report comparison guide explains duplicate reports and overlapping coverage before you combine files into a larger investigation.

Distinguish alignment failures from authentication and handling

Read policy_evaluated SPF and DKIM as the receiver’s reported alignment outcomes. Read auth_results as the separate raw authentication declarations, including their associated domains and any selectors. A raw SPF pass for a service’s envelope identity does not automatically establish alignment with the author domain. Similarly, a validly reported DKIM pass for an unrelated signing domain is not enough to substitute a DMARC pass. The analyzer preserves the declared relationships; it does not reconstruct them with live queries. Consult the SPF and DKIM alignment guide before asking a sender administrator to change configuration based on the word pass in the wrong part of a row.

Disposition is another axis. Values such as none, pass, quarantine, reject, or another reported value describe handling declarations, not a replacement for aligned method outcomes. A row can fail the derived alignment summary while reporting no restrictive handling, for example because the receiver recorded an override or the reported policy context did not request enforcement. Preserve any raw reason and comment as declarations, not verified explanations of the entire route. A quarantine disposition is not proof that a user saw a spam-folder message, and reject is not a measurement of every delivery attempt your organization made. Keep policy, alignment, and disposition in separate sentences when explaining a finding.

Consider an approved invoicing application whose rows report raw authentication passes but both aligned methods failing. The practical question is whether that application was configured to use the intended author, envelope, and signing identities in the affected period. It is not whether every passing token can be counted as a success. In another case, aligned DKIM may pass while aligned SPF fails; the derived result still passes, although the SPF difference may merit operational review. Preserve both method outcomes so a healthy mechanism does not hide a separate change. This tool performs no SPF evaluation, DKIM cryptographic verification, or fresh DMARC policy decision on the original messages.

Follow the count back to the row

Inspect the reported source evidence

Keep the report context, aligned outcomes and raw declarations together while reviewing an unfamiliar source.

Compare sources with an approved sender inventory

Build or consult an inventory outside the analyzer that identifies approved sending services, business purposes, responsible administrators, intended author domains, and expected envelope and signing identities. Include the time when each service was active and any documented route changes. An inventory based only on remembered IP addresses will become stale when a service changes infrastructure or uses shared capacity. The useful match is an evidence-backed relationship between a business-approved mail stream and the reported context. Record where that relationship came from, such as an internal configuration record or a service administrator’s confirmation, without placing account credentials or unnecessary customer details in your working notes.

Divide the review into known and not-yet-explained sources without treating either category as a safety verdict. A known source can still have an alignment problem, an obsolete configuration, or a use outside the approved purpose. An unfamiliar address may belong to an authorized route that your inventory missed; the report alone does not establish that explanation either. Mark the uncertainty explicitly and assign the follow-up to the team responsible for the relevant author domain. Avoid labels that assert attacker, owner, compromise, or legitimacy from the address alone. An approved vendor name mentioned by a colleague is a lead to corroborate, not permission to relabel every row from shared infrastructure.

Ask targeted questions that someone can answer from administrative evidence: was this application active in the window, which author domain did it use, did its envelope or DKIM configuration change, and do its logs show matching activity? Request only the necessary evidence through an approved channel. The analyzer performs no ASN, PTR, IP reputation, geolocation, or DNS lookups and does not infer an operator from an address. Do not add those conclusions to its output as though they were measured here. Where the source remains unexplained, retain an unresolved finding with its count and context instead of forcing a binary conclusion just to close the spreadsheet or support ticket.

Inspect row details before proposing a correction

Use the report filter to narrow the investigation, inspect the source-IP grouping, then open the relevant row detail. Confirm that the count, author domain, aligned outcomes, disposition, and raw authentication declarations belong together. A source summary can combine rows with materially different results; its biggest number does not explain every constituent row. Open the plain raw XML viewer for the selected report when you need to check the supplied wording or an unfamiliar policy option. Reported legacy pct and newer np, testing, and discovery_method values remain report context, not evidence of the domain’s current configuration or a fresh assessment by this website.

For a recognized sender, write a narrow remediation request tied to the affected rows and period. For example, ask the service administrator to compare the intended author domain with the configured signing identity for a specific application. Avoid prescribing a broad policy relaxation merely because some volume failed. Changes should be evaluated by the authorized mail team using its own configuration records and tests outside this tool. If a route changed midway through the reporting window, an aggregate count cannot tell you the exact minute each instance occurred. Preserve that uncertainty and seek appropriately scoped logs rather than constructing a detailed timeline the report does not contain.

A question about one delivered message needs different evidence. The Email Header Analyzer can help read supplied headers from that single message; it does not join the message automatically to an aggregate row or validate signatures. Compare authorized evidence manually and keep the limits of each source clear. A matching domain and approximate time do not prove that a particular message contributed to a particular aggregate count. Keep any sensitive original under your handling rules, and share only the minimum necessary information with the administrator. The DMARC analyzer provides no automated redaction, Copy action, export, PDF, saved project, persistent storage, or investigation history.

Finish with a bounded finding and a cleared session

Prepare inputs within every limit: up to 10 files, at most 4 MiB each, and 16 MiB total source; each expanded XML is at most 4 MiB, with 16 MiB expanded across the batch. Reports are limited to 5,000 rows each and 10,000 rows per batch, XML depth 24, 150,000 nodes, and 4,096 UTF-16 units of leaf text. All limits apply together. Only UTF-8, with an optional UTF-8 BOM, is accepted. The XML, GZIP and ZIP guide covers the supported legacy and RFC 9990 namespaces, strict archive rules, and structural validation boundaries before you retry a rejected input.

Analysis runs locally in an outer worker with a 12-second deadline. A failed input rejects the batch atomically; a successful-looking subset is not published as partial success. There is no API, mailbox connection, report upload, domain lookup, or external entity retrieval. DTD and entity declarations and external processing instructions are rejected. This is a bounded structural reader rather than full XSD certification. A successful parse makes the supplied declarations easier to investigate, not inherently trustworthy. Keep the final finding limited to what the accepted reports say and what your separate administrative evidence actually corroborates, with unresolved questions left visible for the next reviewer.

Clear stops active work and removes the input, file selection, statistics, filters, row details, raw XML viewer, and all derived results. A hidden page, pagehide, pageshow, or freeze also clears the working state; returning to the page does not restore the investigation. This is not secure memory erasure or protection from extensions, a compromised device, or a hostile parent replacing the interface. Handle any external notes and originals separately. Return to the DMARC Report Analyzer overview for another report-reading task or browse Tools without attaching report data to a navigation link.

Questions about DMARC Failures and Source IPs

Does an unknown source IP mean someone attacked my domain?

No. It means the source is not yet explained by your evidence. Compare its report context with an approved sender inventory and ask the responsible administrator for corroboration. The analyzer neither identifies the operator nor assigns an attacker label, and a known source still needs its alignment results reviewed.

Why can raw SPF or DKIM pass while the DMARC summary fails?

Raw authentication and alignment answer different questions. The derived summary uses reported aligned SPF and DKIM: either pass gives pass, both fail gives fail, and other combinations remain unknown. A raw pass for another identity does not replace the receiver’s aligned result or trigger a new verification here.

Can I interpret the failure total as undelivered emails?

No. Counts are reported message instances associated with rows, not unique emails or a delivery ledger. Keep disposition separate from alignment, retain report and window context, and investigate actual delivery questions with authorized receiver or sender evidence outside the analyzer.

Continue with another DMARC-report guide

Read an aggregate report in the DMARC Report Analyzer & XML Viewer or choose a related task below. Explore more practical utilities in all free tools.

A focused investigation starts with context

Review failures without guessing ownership

Analyze authorized reports locally, check the relevant rows, and clear the working session when finished.