About this article: This article is reviewed against the cited sources. This is source research, not lab validation. Check the stated versions and your own requirements before making changes.
Windows records what happens on a server in event logs, and each log is a file of limited size. When the file is full, Windows normally overwrites the oldest events. So how far back the logs reach is not a fixed number of days; it depends on the log size and how fast events arrive. On a busy server, the Security log can cover days, not months (our reading).
The Australian Signals Directorate (ASD) says that the Security log's default size is 20 MB and that this needs to be increased, and that events should be forwarded to a separate, secured place so that an intruder or a full disk cannot remove them. This guide explains the mechanics, the decisions to agree with whoever runs the server, and a few read-only checks. It is general information, and it does not state any legal retention period.
Why it matters
An investigation after an incident asks questions such as: when did the first unusual sign-in happen, which account was used, what changed, and what was touched? The answers sit in logs. If the logs have been overwritten, those questions may have no answer, and the cost is more than curiosity: it affects what you can tell your customers, insurer and regulators, and how sure you can be that the problem is fixed (our reading).
ASD's joint guidance with international partners puts it directly: "Default log retention periods are often insufficient." It adds that "In the event of a cyber security incident, an absence of historical event logs will frequently have a negative impact on cyber security incident response activities." It also points to the length of time it can take to notice a problem, saying that in some cases it can take up to 18 months to discover an incident (Best practices for event logging and threat detection, published 22 August 2024).
Logs are not backups. A backup is a stored copy of data you restore from. A log is a record of what happened. A backup of a server may contain old log files, but only whatever was in them when the backup ran, so it does not replace deliberate log retention (our reading). Replication and failover copy a running system, including any gaps in its logs.
How Windows keeps its logs
Each Windows event log has a maximum size and a retention mode. Microsoft's description of the wevtutil tool sets out the two behaviours when a log reaches its maximum size: "If an event log reaches its maximum size and the log retention mode is true, existing events are retained, and incoming events are discarded. If the log retention mode is false, incoming events overwrite the oldest events in the log." (wevtutil, updated 3 February 2023).
Put plainly:
- Overwrite oldest (the usual setting): the log always holds the newest events, and history is lost from the back.
- Retain and discard new: the log stops recording once full. Old evidence is safe, but new events are lost, which can be worse during an attack (our reading).
Neither setting solves the problem alone. What decides how far back you can see is the size, the event rate and any copy made elsewhere.
ASD's Windows guidance states the default sizes in its own terms: "The Windows default settings have log sizes set to a relatively small size and will overwrite events as the log reaches its maximum size." It says "the Security log size needs to be increased from its default size of 20 MB". The Microsoft pages we read do not state default sizes, so the 20 MB figure is ASD's statement for the Windows versions it covers; check your own server.
How far back do the logs go?
There is no standard answer. Our reading is that retention in time is roughly the log size divided by the amount of event data written per day. Both numbers are specific to your server. A server that records many sign-ins and process events fills a small log quickly, and a quiet one may keep months.
The following figures are invented. If a Security log is limited to 20 MB and the server writes 5 MB of security events a day, the log covers about four days. If the same server writes 5 MB a day to a 2 GB log, it covers about 400 days. Real rates vary with audit settings, the number of users and attacks such as password guessing, which can fill a log fast and push older evidence out.
That last point is a risk worth naming: noisy events can overwrite quiet, important ones.
Larger logs are not enough
ASD's Windows guidance gives an example configuration: it provides Group Policy settings that raise the maximum Security log size to 2 GB and the Application and System logs to 64 MB, as a balance between data use, local retention and performance. It notes these changes increase storage needs on each host.
Local logs have a second weakness. ASD's event-logging guidance says: "some malicious actors are known to modify or delete local system event logs to avoid detection". Windows itself records when the Security log is cleared: Microsoft's documentation for event 1102 says "This event generates every time Windows Security audit log was cleared" and its example includes the account that did it (event 1102). That record sits in the same log an intruder with enough rights can clear, so a copy kept elsewhere is what makes the evidence dependable.
ASD recommends forwarding: "Forwarding event logs to a centralised and secure storage capability prevents the loss of logs once the local device’s storage is exhausted." It also says to store them "in a separate or segmented network" with protection from unauthorised changes and deletion, and to back them up.
How forwarding works on Windows
Windows Event Forwarding (WEF) is a built-in way to send events from servers to a central collector. Microsoft describes it this way: WEF "reads any operational or administrative event logged on a device in your organization and forwards the events you choose to a Windows Event Collector (WEC) server" (Use Windows Event Forwarding to help with intrusion detection, updated 18 August 2025).
Two details from that page matter to a buyer:
- WEF depends on the local logs being set up properly. Microsoft states: "WEF is a passive system regarding the event log." It cannot change the size of log files, enable disabled channels, change permissions or adjust the audit policy. If the server is not recording an event, forwarding cannot supply it.
- By default, "WEF subscriptions will only forward events generated after the WEF subscription was received by the client". So a subscription set up after an incident forwards only what happens from then on, unless it is configured otherwise. Microsoft says that for its Targeted subscription, "enabling the "read existing events" option should be set to true to allow collection of existing events from systems". That means events still held in a server's local log can be collected, but forwarding cannot bring back events that have already been overwritten, deleted or never recorded. Whether a given provider has done this is a question to ask, not an assumption.
The collector side is described in Windows Event Collector (updated 14 July 2025). In a source-initiated subscription, the collector defines the subscription and source computers are set up, using a group policy setting, to forward to it. In a collector-initiated one, the collector lists the source computers.
WEF is one option. Many environments use an agent that ships logs to a log service instead. Either way, the decisions are the same: which events, where they go, who can delete them, and how long they stay (our reading).
What to keep and for how long
ASD's Windows guidance says that, as detailed in the Strategies to mitigate cyber security incidents, "the recommended event log retention time is at least 18 months", and that some organisations may have regulatory requirements for longer. The joint guidance says retention should be informed by an assessment of risk and by any regulatory requirements and frameworks that apply. We do not state any regulatory period here; check the rules that apply to your sector.
A practical split from the joint guidance is hot and cold storage: recent logs that are quick to search, older logs kept more cheaply. Timestamps matter too: the guidance recommends an accurate, consistent time source across systems so that events from different machines line up.
Choosing and agreeing a logging arrangement
The following is our recommendation, not provider guidance.
- Decide who owns logging. On a hosted server, is it you, the provider, or shared? Write down who configures log sizes and forwarding and who can read them.
- Choose a retention target and size the logs and storage for it. Work from event volume, not guesswork, and test it.
- Forward to a separate place. The destination should be outside the reach of the server's own administrators where practical, with deletion restricted.
- Keep the audit policy appropriate. Forwarding cannot supply events that are never recorded.
- Check clock sources. Agree on one time source.
- Watch the logging itself. A forwarding failure or a cleared log should raise an alert.
- Test retrieval. Pick an old date and confirm you can find events for it.
Read-only checks
These checks read settings and events and change nothing. They were not tested for this article. On Windows servers, reading the Security log needs elevated rights; Microsoft says that if PowerShell is not run as an Administrator you may see errors about retrieving log information (Get-WinEvent).
PowerShell shows the size, mode and record count of the Security log:
Get-WinEvent -ListLog Security | Format-List -Property LogName, LogMode, MaximumSizeInBytes, RecordCount, FileSize
The wevtutil command, in a Command Prompt or PowerShell window, displays the configuration of the Security log, including its maximum size and retention setting, and then the oldest event it holds. In that second command, /rd:false reads oldest first, because /rd:true returns the most recent events first, and /c:1 limits the output to one event:
wevtutil gl Security
wevtutil qe Security /c:1 /rd:false /f:text
Compare the date of the oldest event with today. That difference is the history you actually have. Option definitions are in the wevtutil page. Do not use the set-log or clear-log forms of the command during an investigation; they change the log.
Questions to put to your provider
- Who is responsible for event log settings on our servers, and who can change them?
- What is the maximum size of the Security, System and Application logs, and what is the oldest event held today?
- Are logs forwarded anywhere else, and who can delete them there?
- How long are forwarded logs kept, and how can we get them during an investigation?
- Which audit settings are on, and is a change to them reported to us?
- What happens if forwarding stops or the Security log is cleared?
- How would we be told, and how quickly could we get old events?
A worked example (illustrative)
The following is invented. A distributor discovers in November that an attacker used a staff account to read files. It asks its hosting provider for the Security log from August, when it believes the account was first misused. The oldest event on the server is from a week ago, because the log is at its default size and a password-guessing attack filled it. No copy has been forwarded. The distributor can say what happened in the last week but not how the attacker first got in. Afterwards it agrees a larger log, forwarding to a separate store with restricted deletion, and a retention target based on a risk assessment, and it checks the oldest event each quarter.
Read next
Related articles you may want next.
- TroubleshootingReading the sign-in section of a security report: failures, lockouts and what to chaseWhich failed sign-in and lockout patterns matter, which are routine, and how Windows events 4625, 4740 and 4624 help you check a server report.Read article
- TroubleshootingThe first hour of a suspected security incident on a hosted serverWhat to do, preserve and avoid in the first hour of a suspected compromise on a hosted server, before your provider or responders take over.Read article
- TroubleshootingRD Gateway Connection Failures: Certificate, CAP, RAP and Transport ChecksFind why RD Gateway connections fail: check the certificate name, read Gateway events 200/201/300/301/304, test CAP and RAP logic, central NPS timeouts, and TCP 443 / UDP 3391.Read article
Risks and limits
- Default sizes and behaviour differ by Windows version, edition and policy. We rely on ASD's statement of a 20 MB default Security log, which we did not verify on a live system.
- Forwarding and bigger logs cost storage, network and effort. Over-collection also creates noise and may collect personal information, which brings its own obligations.
- Forwarded logs help only if events are being recorded in the first place.
- A forwarded copy can still be lost if an intruder reaches the collector. ASD recommends segmented storage with controls for that reason.
- This article does not cover Linux logs, cloud service logs or log analysis.
Next step
Check the oldest event currently held in the Security log of each important Windows server, and ask who is responsible for sizing and forwarding. If you are reviewing a managed Windows hosting arrangement, the Managed Services page is a starting point for that conversation; ask for the scope in writing.
Sources
- Best practices for event logging and threat detection, Australian Signals Directorate's Australian Cyber Security Centre, published 22 August 2024.
- Windows event logging and forwarding, Australian Signals Directorate's Australian Cyber Security Centre, published 31 July 2017, last reviewed 6 April 2023.
- wevtutil, Windows Server documentation, Microsoft Learn, last updated 3 February 2023.
- Get-WinEvent, PowerShell documentation, Microsoft Learn, date unavailable.
- Use Windows Event Forwarding to help with intrusion detection, Microsoft Learn, last updated 18 August 2025.
- Windows Event Collector, Microsoft Learn, last updated 14 July 2025.
- 1102(S): The audit log was cleared, Microsoft Learn, last updated 7 September 2021.
All accessed 15 November 2026.
Learning notes
- A Windows log is a file of limited size; by default the oldest events are overwritten when it is full.
- How far back you can see depends on log size and event rate, not on a fixed number of days.
- Forwarding copies events somewhere safer. It normally sends only new events, though a subscription can be set to collect events still held locally. It cannot restore events that were overwritten, deleted or never recorded.
- A log is not a backup. Backups do not replace deliberate log retention.
- Check the oldest event in the Security log now, before you need it.
This article was originally published on .