Forensic Case Study: Lockouts Worth Investigating

A lockout flood can be a signal worth checking

A flood of account lockouts is easy to dismiss as a nuisance. This composite case study follows a slow password spray through its days in order. It shows why the quiet days can matter as much as the loud one.

Composite case study: based on patterns observed across incident response engagements; details are blended and illustrative.

The situation

The help desk sees a surge of lockouts across many accounts, and the users insist they typed nothing wrong. The organization’s first move is to treat it as noise. Accounts get unlocked, people are asked to be patient, and the queue moves on.

That is a common first move and a costly one. A lockout can be a symptom of someone guessing passwords at scale. A lockout threshold only catches guesses that cluster together. Where the lockouts originate also matters: Entra smart lockout tracks familiar and unfamiliar locations separately, so a spray from unfamiliar sources often does not lock out legitimate users, while lockouts from on-premises directory or federated sign-in behave differently. Knowing which system is locking accounts tells the investigator which logs to read.

In the following days the lockouts thin out, which feels like the problem fading. In a slow spray it often means the opposite: the guesses have been paced to stay under the threshold and spread out so each day looks unremarkable.

What the investigator saw

The investigator starts with the Entra sign-in log, adding the on-premises directory or federation logs where the lockouts came from there, and does not judge one day alone. Counting failures by source address and by result across the whole window shows what single-day views hide. A set of addresses, which sprays commonly distribute across hosting or proxy infrastructure, tries a small number of common passwords against many accounts and then moves on. Invalid-password failures dominate.

Failure volume alone is not compromise, so the investigator checks the benign explanations alongside the hunt. A security scanner, an internal test or a stale saved password on a phone can all produce failures. Those typically repeat against one account from one device. A spray looks different: one password, many accounts, few repeats.

How it is confirmed

Each question has its own log. Sign-in records show who tried, from where and with what result. Where auditing and licensing capture them, inbox rules, forwarding and mail downloads are found in the Unified Audit Log and mailbox audit, and what can be answered depends on retention and searchability. New authentication methods and app consents are found in the Entra audit log.

What it costs

The cost is mostly time and trust. The quiet days are time the actor can use without being watched. Staff effort goes into unlocking accounts that could have gone into scoping the problem. Trust takes a hit too, because affected users want to know whether their mail was read. The answer depends on logs that are retained and searchable.

Containment

A suspicious success from a spray source is contained promptly, without waiting for full confirmation. Confirmation and evidence capture run in parallel and are not a reason to wait.

1. Reset the password for the account with the suspicious success, or disable it, then revoke sessions and refresh tokens immediately after, in the same sitting, and re-check for new sessions. In a spray the attacker already holds the password, so a revoke without a reset can let them sign in again.

2. Remove rules, forwarding, authentication methods or app consents the session created.

3. Export the relevant sign-in and audit records while containment proceeds.

4. Block the sources seen in the spray, understanding that this is a stopgap because addresses commonly rotate.

5. Hunt for other quiet successes, starting with dormant and rarely used accounts.

6. Reset passwords for accounts that succeeded from spray sources, require a password change where needed, and require MFA broadly. A custom banned-password list only affects future password changes and does nothing to weak passwords that already exist. Review smart lockout settings and pair these steps with blocking legacy authentication and phishing-resistant MFA.

Lessons

  • Lockouts are data. A burst of them across many accounts deserves a look at the source, not just an unlock.
  • Review a spread-out window, not a single day. Slow activity is designed to look like typos.
  • A single hit can be one of several, so scope the whole window rather than stopping at the first success.
  • Older protocols can let a guessed password work without a second factor, so check which protocols a suspicious session used.

Checklist for your next incident

  • Count failures by source and by result across the full window.
  • Look for a success from any source that failed against other accounts.
  • Contain a suspicious success promptly: reset or disable the account, revoke sessions and refresh tokens right after, then re-check for new sessions.
  • In parallel, confirm with the user, device, location and session activity, and rule out scanners, tests and stale saved passwords.
  • Review what each successful session did next: rules, forwarding, downloads.
  • Check dormant accounts before closing the case.
  • Review smart lockout, block legacy authentication and require phishing-resistant MFA. A banned-password list only affects future changes, so also find and reset existing weak passwords.

Message us to talk through your password spray response plan.


Discover more from Core-Infosec

Subscribe to get the latest posts sent to your email.

Picture of Core Watch

Core Watch

Practical security guidance from Core-Infosec.

Read More