After a shared login changes hands, check user changes, group changes, the review window, and recorded activity to establish who held access and when.
Key takeaways
- Match user and group identifiers to the machine account.
- Compare recorded changes with approvals.
- Record the review dates and when you pulled the log.
- Flag unexplained activity and gaps before summarizing access.
The answer has to come from records. A recollection of who was supposed to have access may not match what the system actually recorded, and the gap between the two is exactly what a reader will ask about.
These four checks use Bitwarden's event logs as the example. Bitwarden describes its logs this way: "Track your organization's activity and investigate incidents with event logs, timestamped records that capture changes and usage across your Teams or Enterprise organization." If your own tool keeps a log, the same thinking applies, but first confirm what that log records and how far back you can read it. Do not assume that another product records the same events, or records them in the same way. A shared login is also harder to review than an individual account, because more than one person may have used it, so the question is about the account, not about one user.
Before you open any log, write down three things: the name of the shared login or machine account, the date the change happened, and the list of people who should have held coverage before and after it. In this post, coverage means the period when a named person could use the shared login or machine account. With those three items in hand, the review becomes a comparison you can repeat, and a comparison is easier to explain than a memory.
A simple worksheet helps. Make one table with a row for each event you find, and columns for the date, the person or group named in the event, the event type, whether the change was expected, and the approval that supports it. Include notes explaining each finding. A summary for leadership does not have to include the raw log, but every line in it should trace back to an event in the log. For another record-review task, see close date change logs.
Who was added to or removed from the account?
Bitwarden records each addition and removal of a user on a machine account as its own event, so you can list every user change inside your review window. Bitwarden documents these two user event types:
- "Added user: user-identifier to machine account with identifier: machine-account-identifier ( 2300)"
- "Removed user: user-identifier from machine account with identifier: machine-account-identifier ( 2301)"
Each event carries a user identifier and a machine account identifier. Those two identifiers are the keys you match on, so copy them into your worksheet exactly as they appear. Do not match on a person's name alone, because the name in your notes and the identifier in the log may not be written the same way. Sort the events by date and mark each one as expected or unexpected. An expected change has an approval you can point to, such as a ticket or a written note from the account owner. An unexpected change has none, and it goes on the list of items that need an explanation before the review is finished. Keep a copy of every event you cite, with the date it appears in the log and the identifiers it shows, so a reader can check the citation later.
Illustrative example: compare approval and removal
Compare the approval's end date with the removal event's date and flag the intervening period for explanation. Your write-up should give both dates and the event code, without treating the later removal as evidence that the extra time was approved. The same logic applies in reverse: if the removal came earlier than the approval described, note that too, since it shows coverage ended sooner than planned.
Did group membership change along with the users?
Yes, if the log shows it. Bitwarden records group changes as their own events, separate from user changes, so check both lists before you call the review finished. Bitwarden documents these two group event types:
- "Added group: group-identifier to machine account with identifier: machine-account-identifier ( 2302)"
- "Removed group: group-identifier from machine account with identifier: machine-account-identifier( 2303)"
Use the group identifier and the machine account identifier as your keys, and mark each group change as expected or unexpected, the same way you handled the user events. Put the group rows into the same worksheet as the user rows. A reader should be able to see every change to the account in one place, not spread across two documents. In the write-up, give group changes their own short section with their own event codes, so a reader can tell which lines came from user events and which came from group events.
If the user events and the group events seem to disagree, do not pick the version you like better. Write down both, mark the disagreement as an open item, and note what you would need to resolve it, such as the approval record or an answer from the account owner. An open item shown plainly is more useful to a reader than a tidy table that hides it.
How far back does the log reach, and when do events appear?
Bitwarden keeps event log data indefinitely, but you can only view up to 367 days worth of data at a time, so a review that reaches further back has to be done in more than one view. Set your window to begin shortly before the first change you found and to end after the last one, then write both dates at the top of the worksheet. A reader will want to know exactly what period you covered, and the window is the answer to that question. For a separate system's limits, see audit log retention checks.
If the shared login or machine account is still in use, the window stays open. Decide on an end date for the review and say so in the write-up, because activity after that date is outside what you checked. Do not present the review as covering the period up to today unless you actually pulled the data for that period.
Timing matters too. Bitwarden says server events are recorded instantly. Client events, which it describes as the majority, are transmitted to the server every 60 seconds. That means a client event that happened a moment ago may not show in the log yet. If you run the review right after a change, pull the log again later and compare the two pulls. If the second pull shows events the first one did not, record that in the worksheet, along with the time you pulled each one.
What did the login or account do during the window?
Bitwarden's event logs capture usage as well as changes, so the same log can show activity during the window, not only the list of people who were added or removed. Bitwarden records over 60 event types, and each one is listed with its type code, so go through that list before you filter. Choose the types that answer your question, and write down which ones you chose and why. A reader who asks why certain events were left out should find the answer in your notes.
For each person on your expected list, look at activity inside their coverage dates and flag anything outside them. Flag activity from an identifier that matches nobody on the list as well, because it needs an explanation before you can say who held coverage. Do not guess at that explanation. If the log does not show who was behind an event, write that the event could not be tied to a named person.
Expected activity is activity that falls inside an approved person's coverage dates and fits the purpose of the approval. Unexpected activity is anything else. Keep the two groups separate in the worksheet, so a reader can see how many events you examined and how each one was classified.
Activity from a person after their removal date needs its own line. Write the event, its date, and its identifier, and state whether it falls after the removal date you recorded in the worksheet. Keep it out of the general list, so it stands on its own in the write-up.
You can access these logs through the web app and the /events endpoint of the Bitwarden Public API. Whichever route you use, record which one you used and the date you pulled the data, so a later check can reproduce the same view.
What should the write-up contain?
The write-up should contain the window you reviewed, every event you relied on, and one line for each person who held coverage. Each line should say when that coverage started, when it ended, and what approval supports those dates. Keep the event type codes and the identifiers in the write-up, not only the plain names. A reader who wants to check your work should be able to find the same event you found, and the codes and identifiers make that possible.
Be plain about gaps. If the log does not cover a period you needed, say so rather than filling the gap with an estimate. If the window was longer than one view could show, say that the review came in parts, and list the parts.
Do not describe the review as proof of more than the log records. A line such as "the log shows no events for this person after the removal date" is accurate. A line such as "this person did not use the account after the removal date" goes beyond what the log supports, unless you have other records that show it.
Give leadership a short summary at the top, and put the worksheet below it. The summary answers the question leadership asked, and the worksheet shows how you know the answer. A line in the summary might read: "Person A held coverage from the start date to the end date, under the approval in the ticket. The log events for that period are listed in the worksheet." Keep the summary to what the worksheet supports, and nothing more.
Why run the checks in this order?
Run them in this order, because each check narrows the next. The user changes tell you who to look for. The group changes tell you whether they add anything to the picture. The window tells you what period you can show. The activity review tells you what the login or account was used for during that period. Running them out of order means you may review activity for a person whose coverage you have not yet established, and then you have to go back and redo the work.
Finish each check before you start the next one, and mark any check you could not finish. An unfinished check is an open item, and open items belong in the write-up.
Today, choose a changed account and compare its latest user-change event with the approval record; note any mismatch before expanding the review.
What if a change was not authorized?
Refer an unauthorized change to the people who handle security incidents in your organisation. If your organisation is in the UK, the National Cyber Security Centre lists resources for individuals and organisations in the UK who have experienced an online scam or cyber attack. Those resources are the NCSC's, not Bitwarden's.
Before the summary goes to leadership, ask the account owner to read the worksheet and check it against their own records. Note their response in the notes column, including any disagreement, so the reader sees it.
Revenue leaders do not need to become log experts to produce that list. They do need to know which checks were run, which events were used, and which questions are still open.
Related guides: Deal change records: 3 tools and their limits · Call log timestamps: 7 checks for fractional seconds.