Sysdig Documentation

[Beta] Activity Audit

Note

From Sysdig Secure 3.0 (with Sysdig Agent 0.93.0+), the Commands Audit feature has been replaced by the more powerful and intuitive Activity Audit module.

Contact Sysdig Support to have Activity Audit enabled for your SaaS Sysdig Secure environment.

Activity Audit takes the high-value data from captures and makes it always-on, searchable, and indexed against your cloud-native assets. This stream includes executed commands, network activity, and kube exec requests to the Kubernetes API.

activity_audit.png

Understanding how Activity Audit is Used

Activity Audit allows users to view different data sources in-depth for monitoring, troubleshooting, diagnostics, or to meet regulatory controls.

Troubleshooting

A system investigation may be triggered by an event generated by Sysdig, or by an alert from another tool or person.

  • Find contextualized, relevant data Activity Audit allows easy access to the underlying data to help trace the event, evaluate its impact, and resolve the issue.

    From Policy Events in Sysdig Secure, jump directly to the relevant Activity Audit to investigate details.

    audit_jump.png
  • Trace commands and connections back to users Activity Audit can correlate the interactive requests from a Kubernetes user with the commands and network connections performed inside the container, allowing an operator to trace this activity back to a user identity.

Regulatory

The Activity Audit can also provide data about the infrastructure to help prove to auditors that proper data visibility and security measures are in place

Activity Audit is a critical requisite for many compliance standards:

Navigate the Audit Interface

Activity Audit displays a continuously updated list of activities. Use the UI features to find and filter the information you need.

audit_panels.png

Groupings

Filter activities by pre-defined groupings from the drop-down menu.

Note that each element in a group shows a summary of the activity entries for that entity. You can use this number to discover areas of high activity; filtering changes the number.

Data Sources

Use the top-right drop-down to filter information from one particular data set. Current data sources are:

  • User commands

  • Network connections

  • Kube exec commands

Frequency Graph

The graph shows the activity frequency for each data source, allowing users easily to zero-in on anomalies.

audit_spike.png

The image above shows a spike in network activity (orange line) between 7:00 - 8:00 pm.

Drag the mouse over the peak to auto-zoom on the time frame and see more detail.

audit_spike_detail.png

Time Navigation Buttons

Use the time window navigation bar to show only activities run within that window. (For more information, see also Time Windows.)

Note

Sysdig Secure does not currently provide the functionality to configure a custom time window.

Detail View

Select an activity row to see its details, including the attributes by which you can filter (whitelist/blacklist).

See Review Activity Details for the attributes of each data source.

You can also use the whitelist (+)/blacklist (-) attribute filters from the detail view.

audit_detail.png

Whitelist (+) and Blacklist (-) Attribute Options

Next to attributes in the Detail View are +/- signs used for filtering.

Click + to include an attribute; click - to exclude an attribute from a filter.

Trace Button (kube exec activity) trace_button.png

Beside each kube exec item is a Trace button trace_button.png.

Using this feature you can correlate container activity to the original Kubernetes user and IP. See Follow a kubectl exec Trace.

Filter

Filtering is the heart of Activity Audit's power. Filters allow you to search, sort, parse, and surface meaningful data and connections as they are needed.

Ways to filter Activity Data:

  • Data Source: Choose a data source from the drop-down: network activity, commands, kubectl exec.

  • Attribute (+/-): Choose + or - next to an attribute to include/exclude that attribute from the filter

  • Attribute (manual): If you know the attribute, you can type it into the filter box manually, with the following syntax:

    Include an attribute

    attribute_name="attribute_value" e.g. comm="grep"

    Exclude an attribute

    attribute_name!="attribute_value" e.g. comm!="grep"

  • Trace Trace kube exec entries to see all relevant activity in that session from that user

  • Time graph: Select a section of the graph to zoom in on a time frame and see detailed activity

  • Combine: These methods can be combined as needed.

    For example, the filter below surfaces activity on a particular pod, while excluding activity from one IP address that is known to be normal.

    resource="pods" name="woocommerce-6877958" sourceaddresses!="172.20.41.2

    audit_filter_example.png

Review Activity Details

Command Details

Name

Description

When

The date and time the command was executed.

Command

The command executed.

Full Command Line

The complete command, including all variables/options.

Working Directory

The directory the command was executed in.

Scope

The entities within the infrastructure impacted by the command.

Host

The hostname and MAC address of the host the command was executed on.

Container

The container ID, container name, and image that the command was executed on.

Additional Details

Detailed user/host information:

  • The Process ID (PID) of the command.

  • The Parent Process ID (PPID) of the command.

  • The user ID of the user that executed the command.

  • The Shell ID.

Network Connection Details

Name

Description

Time

The date and time of the network connection

Connection Direction

Incoming or outgoing connection

Connection Details

Including:

  • Transport-level protocol (lp4)

  • Client address, server address (lp4)

  • Client port, server port

Scope

The entities impacted by the network connection

Host

The host name and MAC address of the host where the connection was made

Additional Details

The process name and ID (Parent Process ID/PID) that launched or received the network connection

Kubectl Exec Details

Name

Description

Time

The date and time of the kubectl command

Kubernetes resource

Including:

  • resource: The kind of Kubernetes resource affected (currently only pods)

  • name: name of the resource (pod name)

  • subresource: currently exec

  • command: the command executed

  • container: the high-level name in the Kubernetes defintion

Kubernetes user and group

Including:

  • user: user name performing the kubectl command. Can be either a service account or a human user.

  • groups: groups the user belongs to

  • userAgent: client userAgent

Sources addresses

External IP address that initiated the connection

Scope

Including

  • Kubernetes namespace name

  • Kubernetes deployment name

  • Kubernetes pod name

  • Container ID

Host

Host name and MAC address of the host where the kubectl exec was made

Sample Use Cases

Blacklist: Commands

I have a noisy command happening every 0.2s, which I know is totally normal in my environment. There are other commands that look more suspicious to me:

Foo
Foo
Foo
Foo
Foo
Suspicious command - curl
Foo
Foo
Foo
… 200 Foo
Suspicious command  - dpkg
…. 500 Foo
Suspicious command  - shred

This is a clear case for blacklisting "Foo" and concentrating on the rest.

Filtering for Incident Response

A Policy Event reports a dangerous peak in network connections coming from a specific pod. This example describes one way to search for the root cause.

What user and what activity triggered this issue?

  1. Use the Activity Audit button next to the policy event to jump directly to the relevant audit trail.

    audit_ex2.png
  2. Here one can determine at a glance:

    • The pod/namespace on which the heightened activity is occurring (where the 3396 network entries are)

    • The process related to the activity (in this case, ab, or the Apache Benchmark tool)

    • Related activities in the graph (cmd and kube exec lines)

    • Repetitive entries that can be screened out

  3. Refine the view through filtering:

    audit_ex2b.png
    1. Switch from network data source to cmd and kube exe.

    2. Filter out noisy, repetitive entries (e.g. comm!="bash")

    3. Investigate details of a kube exec item for user information.

  4. After filtering, you have a focused incident report detailing:

    • The Kubernetes user "johndoe"

    • The external IP he used to connect

    • The set of commands he used to install and launch the Apache Benchmark stress-testing tool.

Follow a kubectl exec Trace

In a production environment, kubectl exec commands are typically suspicious. Also, because such commands are interactive sessions, it can be difficult to pinpoint which individual has issued the command(s) and what other activities the individual performed. This is where Sysdig's Trace functionality comes in, correlating kubectl exec commands with a specific user and the network and command activities performed in that user's session.

In this example, suspicious activity has been detected and you want to determine whether someone has downloaded and executed a Trojan horse.

  1. Use the Groups to display your Kubernetes hierarchy by namespace and deployment. Focus on the pod displaying unexpected high levels of activity (based on the number in parentheses).

    trace1.jpg
  2. Checking the corresponding activity graph, you zero in on a time frame and see kube exec activity among the hundreds of commands and network events.

    trace2.png
  3. Select the kube exec item and click the Trace button on left.

    This session trace will display a formatted report of any container activity (network, commands) that the user performed inside the container.

    trace3.png