Explore our new Cybersecurity Terms Glossary, explained by our experts.
Cybersecurity Terms Glossary
Authentication log analysis is the examination of sign-in records to find credential abuse. It depends on knowing which event codes, logon types, and failure reasons actually distinguish an attacker from a user.

What Is Authentication Log Analysis?
Authentication logs record who proved their identity, from where, by which method, and whether it worked. Because most intrusions now involve valid credentials at some stage, these records are where a great deal of detection either happens or fails to. The difficulty is volume and ambiguity: a large estate produces millions of sign-in events a day, almost all of them legitimate.
On Windows, 4624 records a successful logon and 4625 a failure, but the useful detail sits in the fields rather than the event number. The logon type tells you the mechanism, and attackers leave distinctive patterns:
| Type | Meaning | Why it matters |
|---|---|---|
| 2 | Interactive | At the console; unusual on a server |
| 3 | Network | File shares, remote tooling, lateral movement |
| 9 | NewCredentials | Classic pass-the-hash signature with runas /netonly |
| 10 | RemoteInteractive | RDP, including attacker hands-on-keyboard |
On failures, the status and sub-status codes separate noise from intent. 0xC000006A is a wrong password for a valid account, 0xC0000064 is a username that does not exist, and 0xC0000234 is a locked account. A burst of 0xC0000064 is enumeration; a slow trickle of 0xC000006A across hundreds of accounts with the same password is spraying, which no account lockout policy will catch.
Kerberos adds its own: 4768 and 4769 for ticket requests, with encryption type 0x17 on a service ticket being the long-standing signal for Kerberoasting against an account that should be using AES. Event 4648, a logon using explicit credentials, is one of the most underused detections in Active Directory, because it exposes credential use that does not match the logged-on user.
Microsoft Entra ID and Google Workspace produce richer records than Windows ever did: result code, conditional access policy evaluation, device compliance state, application, client app type, and risk assessments. Three patterns are worth building on. Legacy authentication protocols succeeding anywhere, because they bypass multi-factor authentication. Sign-ins where multi-factor was satisfied by a token rather than a fresh challenge from an unfamiliar location, the fingerprint of session theft. And a failure sequence that resolves into a success from a different address within minutes, which is what credential stuffing looks like when it works.
Retention is the practical constraint. Entra sign-in logs are kept for 30 days by default, shorter than the time most account compromises take to surface, so forward them into your SIEM and treat that as a retention planning decision rather than an afterthought.
Every one of those detections is a statement about what is normal, and normal differs per account. A 24-hour pattern for a UK-based finance team is genuinely different from one for a globally distributed engineering group, and a single threshold applied to both produces noise in one and blindness in the other. Build per-entity baselines over at least 30 days, and store the reason for every exclusion you add, because exclusion lists grow and nobody remembers why the third entry is there.
At Privia Security, identity telemetry is the first thing we assess in our SOC maturation services, because detections built on anything else tend to assume that authentication is already understood.
Event 4624 for successful logons and 4625 for failures are the core, read together with the logon type and the failure status code. Add 4648 for logons using explicit credentials, 4768 and 4769 for Kerberos ticket requests, 4672 for assignment of special privileges, and 4776 for NTLM validation. On domain controllers these are high volume, so collection and retention need planning before detection.
Lockout policies count failures per account. Spraying attempts one or two common passwords against many accounts, so no single account reaches the threshold. Detection has to look across accounts instead: count distinct usernames failing from the same source within a window. Set the alert on distinct accounts rather than total failures, and remember that cloud sign-in logs and on-premises logs need separate rules.
On its own, no. VPN egress points, mobile roaming, cloud-hosted email clients, and geolocation errors all generate false matches, and a determined attacker using a residential proxy in the same city will not trigger it at all. It works well as a contributing signal in a risk-based model, combined with a new device, a new MFA registration, or a mail rule change.
Twelve months is a sensible target, with the most recent 90 days immediately searchable. Account compromises are frequently discovered weeks after the initial sign-in, and investigators need to establish the first access, not only the most recent. Cloud platforms often default to 30 days, so forwarding to a SIEM or archive tier is usually required rather than optional.
You May Be Interested In These
We use a few cookies that are needed to make this website work, and they are always on. We would also like to set analytics cookies, to see which pages are useful; functional cookies, so that our live chat works; and advertising cookies, to measure our campaigns. We will not set any of those until you say yes, and you can change your mind at any time from the Cookie Settings link in the footer.
Choose the cookies you are happy for us to use. Strictly necessary cookies are always on; everything else is off until you turn it on.
These cookies keep the website working: they hold your security checks, your form submissions and the choice you make here. The site cannot run without them, so they are always on.
These let our live chat open and remember your conversation, and let the newsletter sign-up work. Switch them off and the chat window will not appear.
These tell us which pages are read and how people find us, so that we can improve the site. The reports are about groups of visitors, never about you personally.
These let us measure our advertising and show you our adverts on other websites, such as LinkedIn, Google and Meta. Switch them off and you will still see adverts, but they will be less relevant.