The two CISA log controls are CPG 2.T, analyzing logs to determine the impact of events and whether an incident occurred, and CPG 2.U, setting up centralized log management using a SIEM tool. Together they tell you what happened and where the proof lives.
The two CISA log controls, answered first
The two CISA log controls for your audit records are analyzing logs to determine what an event affected and whether an incident has occurred, and setting up centralized log management with a security information and event management (SIEM) tool. Together they answer the questions you will face after something goes wrong: what happened, and where is the proof.
CISA's #StopRansomware Guide lists both controls with its own reference codes, CPG 2.T for log analysis and CPG 2.U for centralized log management. Those bracketed codes are the guide's identifiers, and they are useful when you ask an IT provider to confirm that each control is in place.
For a small accounting firm, the audit records that matter most show who opened a client file, who changed it, and who exported or shared it. Staff chat records matter too, because a client conversation is often where a problem first appears. Logs are the raw trail behind all of this. Without them, you may know that a file was altered but not when, by whom, or from which device.
The two controls work in sequence. Centralizing logs gathers the trail in one place, and analysis reads that trail to decide whether an event was an incident and how far it reached. Neither works well alone. A tool that collects logs nobody reads gives you little, and analyzing logs scattered across several systems takes far longer than analyzing one collected stream.
Chief takeaways
- Treat log analysis and centralized log management as two separate controls, each with its own CISA reference code.
- Map your systems and data flows before deciding what to log.
- Use a security information and event management tool so logs from different systems land in one place.
- Check what your chat tool keeps and deletes, because retention settings decide what survives.
- Bring IT, your managed security provider, your cyber insurer, and firm leaders into the plan early.
Map systems and data flows before logging
Before you decide which logs to keep, you need to know what you are logging. CISA asks organizations to "Develop and regularly update comprehensive network diagram(s) that describes systems and data flows within your organization’s network(s) (see Figure 1) [CPG 2.P]."
For a firm of perhaps ten people, the diagram can fit on one page. It shows the file server or cloud storage holding client folders, the accounting software, the email system, the chat workspace, and the laptops and phones that connect to them. The most useful part is the arrows, which show where client data travels from one system to another. A log that covers only one box misses the handoffs, and handoffs are often where problems hide.
The guide also uses two figures to show network layout: "See Figure 2 and Figure 3 for depictions of a flat (unsegmented) network and of a best practice segmented network." The table below sets out what each of the three figures covers.
| Figure | What it depicts | How the guide presents it |
|---|---|---|
| Figure 1 | Systems and data flows within your network | Develop and regularly update as a diagram |
| Figure 2 | A flat (unsegmented) network | Shown as a flat example |
| Figure 3 | A best practice segmented network | Shown as best practice |
A flat network has no internal divisions. A segmented network separates its parts, so a problem in one area is less likely to reach the others. When you draw your diagram, mark which systems hold client data and which connect to them. Those marks tell you where logs matter most.
Centralized logs with a SIEM tool
The control is direct: "Set up centralized log management using a security information and event management tool [CPG 2.U]." The #StopRansomware Guide names this as the method, and the diagram from the previous step tells you what to connect.
| Log source | What it records | Why it matters |
|---|---|---|
| Staff accounts | Sign-in records | Shows who signed in and from where |
| File storage | Activity logs | Shows who opened, copied, or changed client folders |
| Accounting software | Records of activity | Tracks changes to client financial records |
| Email and chat | Audit or export logs | Captures exports and shared client conversations |
A SIEM tool collects logs from many sources and stores them together, so you can search across them without opening each system separately. For your firm, the sources to connect are the ones on your diagram that touch client data: sign-in records for staff accounts, activity logs from file storage, records from accounting software, and the audit or export logs from email and chat.
Start with the sources that hold the most sensitive client files. Confirm that each one sends its logs to the central tool and that the logs arrive with timestamps and account names attached. A log without a clear time and user is hard to use later, when you need to reconstruct a sequence of events.
Access to the central tool needs the same care as access to client files. Limit who can view the logs, and record who has that access. A central log store is valuable to an attacker as well, because it shows how your systems work. The tool should be configured so that the people who review logs are the people who are authorized to see them.
Test the setup before you need it. Pick one ordinary action, such as opening a client folder in your file storage, and confirm that it appears in the central tool with the right account name and time. If it does not, the gap is in the connection, not in the event. Repeat the test after any software update or new staff account, since a change in one system can stop logs from arriving.
What logs show after an event
CISA states the purpose plainly: "Logs can be analyzed to determine the impact of events and ascertain if an incident has occurred [CPG 2.T]." Log analysis (CPG 2.T) only pays off when someone reads the logs. The #StopRansomware Guide gives analysis two jobs: finding out whether an incident happened and measuring how far it reached.
| Analysis step | Purpose | What it answers |
|---|---|---|
| Pull logs for the affected account or system | Start the investigation | Which records relate to the event |
| Set a time window before the first unusual activity | Bound the search | When the activity began |
| Ask whether the activity was an incident | Apply CPG 2.T | Did an incident occur? |
| Ask which files, accounts, or systems were touched | Measure reach | How far did it reach? |
When something looks wrong, such as a sign-in from an unfamiliar location or a sudden export of many client files, start by pulling the logs for the affected account or system. Set a time window that begins before the first unusual activity. Then ask two questions. Did the activity amount to an incident? If it did, which files, accounts, or systems did it touch?
The second question matters most for a firm. Logs can show whether a client folder was opened, copied, or changed, and when. That answer tells you what to investigate, which records need to be restored, and what you can say to affected clients. Without that answer, you may end up treating every client folder as affected.
Preserve logs before you change anything. If a client folder has been altered, resist the urge to restore or delete files until the relevant logs are copied out of the systems that generate them. A restore that runs before analysis can overwrite the evidence you need. Note who performed each step and when, so the record of your own response is as clear as the record of the event.
Make the analysis routine instead of a scramble. Decide in advance who reviews the logs, how often, and what counts as worth a closer look. A short written checklist makes the review repeatable, so the same questions get answered the same way whether the event happens on a quiet Tuesday or during the busy season.
Slack retention: keep everything or delete on schedule
Staff chat is often the least controlled audit trail in a small firm. Slack's help article on data retention describes two behaviors that matter for your records. The first is "Slack will keep all messages, but will not track message edits or deletions." The second is "Slack will delete messages following the custom time frame you choose."
The table below compares the two behaviors by what Slack does and what that means for an audit trail.
| Setting | Slack's wording | Effect on your audit trail |
|---|---|---|
| Keep all messages | "Slack will keep all messages" | The full message history stays in place |
| Custom time frame | "will delete messages following the custom time frame you choose" | Messages are removed on your schedule |
Keeping everything gives you a complete history, but it does not show whether a message was changed or removed after it was sent. That gap matters when a dispute turns on what a staff member wrote and when. A custom time frame limits how long old conversations sit in the workspace, which reduces exposure. The cost is that messages disappear on schedule, so any record you need must be exported before the window closes.
Choose the setting deliberately, not by default. If your firm needs chat records for client work or for proof of advice, a keep-all setting protects the history. If you want to limit how long client conversations are stored, a custom time frame does that, provided you pull out the records you must keep first.
Org-level policy and workspace limits
Slack's help article on data retention describes a control that sits above individual workspaces: "Org Owners can set a policy for message retention that will apply to every workspace in their org." A small firm with one main workspace may see this as a formality, but it matters once you add a second workspace, a client-facing channel, or a temporary group for a project. Without an org-level policy, each workspace could end up with its own retention rule, and nobody may remember which one applies.
The article also limits local control: "By default, when an org-level policy is set, Workspace Owners can’t adjust message retention settings for their workspace." In practice, the person who handles day-to-day chat administration may not be able to change retention once the org policy is in place. For audit records, that can be a benefit, because a single administrator cannot quietly shorten the history. It can also be a surprise if the only person who knows the settings is a workspace owner who expects to change them.
Before you set an org-level policy, decide who will hold the org owner role. That person should be a partner or someone who reports directly to one, not a staff member who may leave or change roles. Write down the chosen setting, the reason for it, and the date it took effect. A firm that can show when its retention rule changed is in a stronger position than one that cannot.
Which one should you pick?
Choose the retention setting based on what your firm may need to prove later. Slack's help article describes two behaviors that pull in opposite directions. Keeping all messages means "Slack will keep all messages," which protects the history of a client conversation. A custom time frame means "Slack will delete messages following the custom time frame you choose," which limits how long old chat stays in the workspace.
For most small accounting firms, the answer depends on how chat is used. If staff use chat to give advice, agree on deadlines, or confirm client instructions, that history may be needed later, and a keep-all setting protects it. If chat is mostly used for scheduling and informal updates, while client records live in your document system, a custom time frame may fit better, because it trims how much old conversation remains exposed.
Whichever you choose, keep the limit of the keep-all setting in mind. The article says Slack "will not track message edits or deletions," so the history survives but does not show whether a message was changed. If that gap matters, your firm needs a habit of recording client decisions in the client file as well as in chat. Check the current settings before you commit to a different rule for each workspace.
Who must be in the room
The #StopRansomware Guide uses the words "may include," so treat this list as a starting point rather than a fixed roster.
For a small firm, your IT department may be an outside provider or one technical staff member. A managed security service provider, if you use one, can confirm which logs the central tool collects and whether analysis is part of its service. Your cyber insurance company often asks what evidence you hold after an incident, so bringing its questions into planning early avoids surprises. In a firm, the departmental leaders are usually the partners or practice leads who decide which client records matter most and who approve retention choices.
Make a short list with one named contact for each role. Ask each contact three questions: which logs they need, how long records must be kept, and who must be told if an incident occurs. Answers recorded in advance save time when a real event forces quick decisions.
Decide before an event how staff should report a concern. A simple instruction, such as telling a named partner directly rather than posting in a general channel, keeps sensitive details out of the very chat workspace you are trying to audit. Write the instruction into your staff guidance so it applies to everyone, including new hires.
Reporting suspicious activity and least privilege
The Microsoft Learn page describes Microsoft Entra identity settings. The statement names three ideas: reporting, risk-driven remediation, and least-privileged administration.
In your firm's terms, reporting means a staff member can flag a sign-in that looks wrong, and the report reaches someone who can act on it. Risk-driven remediation means the response depends on how risky the activity appears, rather than treating every alert the same way. Least-privileged administration means the people who change sign-in and security settings have only the access their jobs require.
Ask your IT provider three things. Is suspicious-activity reporting turned on for your staff accounts? Who receives those reports? Who holds administrator rights, and does each of them need that level of access?
Ranking the retention choices in one table
The table below ranks Slack's retention options by how completely each one preserves the audit trail, using the wording from the help article.
| Rank | Option | Slack's wording | What it means for your records |
|---|---|---|---|
| 1 | Keep all messages | "Slack will keep all messages, but will not track message edits or deletions." | Full message history remains, but edits and deletions leave no tracked trail |
| 2 | Custom time frame | "Slack will delete messages following the custom time frame you choose." | Old messages are removed on schedule, so exports must be made before the window closes |
| Not ranked | Org-level policy | "Org Owners can set a policy for message retention that will apply to every workspace in their org." | Enforces either choice across workspaces, so it is a way to apply the rule, not a rival to it |
The ranking reflects audit completeness only. A keep-all setting gives the fullest history, while a custom time frame gives less history by design. The org-level policy is not ranked because it applies whichever retention rule you select. If your firm needs both completeness and limited storage, a common approach is to keep records that matter for client work and export them before any custom deletion takes effect.
Worked example: checking one log end to end
A firm of ten people draws a one-page diagram showing cloud storage, accounting software, email, and chat, with arrows marking where client files travel. It connects the storage and email audit logs to its central tool. One Tuesday, an export of many client files appears in the tool with the account name and timestamp attached. The team pulls logs from the hour before, confirms the account and device, and copies the logs out before restoring any files. The check takes a few minutes and answers both questions: the activity was an incident, and one account touched a set of folders.
Related guides: Account deletion for firm records: 3 checks before removal · Audit log retention: 4 settings to check for your firm · Client data access controls: 4 tools for a firm WISP.
Reader questions
Does CISA's guide show what a flat network looks like? Yes. The #StopRansomware Guide refers to Figure 2 as a depiction of a flat, unsegmented network, and it shows Figure 3 as a best practice segmented network.
Which network layout does CISA present as best practice? The segmented layout.
Can a workspace owner change retention after an org-level policy is set? By default, no.
Does an org-level policy reach every workspace? Yes. The article says an org owner's policy "will apply to every workspace in their org," so one decision covers all of them.
Will a keep-all setting show that a message was edited? No. The help article says Slack keeps all messages but does not track message edits or deletions, so you should keep client decisions in your client file too.
What happens to messages under a custom time frame? Slack deletes them following the time frame you choose, as the help article states. Export any records you must keep before that window closes.