auditd daemon configuration:
/etc/audit/auditd.conf

Audit rules:
/etc/audit/audit.rules
/etc/audit/rules.d/

Audit logs:
/var/log/audit.log*
/var/log/audit/audit.log.*.gz

Overview

The Linux Audit system is an non-default auditing and logging framework that can be configured to log multiple type of operations, such as authentication successes or failures, process executions, file accesses, user commands executed in a TTY, etc.

The Linux Audit framework implements kernel-mode hooks to monitor user-mode processes and generate audit telemetry. The auditd daemon is the main user-mode component of the Linux Audit framework. It receives audit messages sent by the kernel (via a netlink socket) and is responsible for storing the messages on the file system. Launched at operating system startup, the auditd daemon is configured through the configuration file /etc/audit/auditd.conf, and connects to the socket to receive events from the kernel. The auditd daemon can also received events from other system components, such as pam or sshd.

Audit rules

The Linux Audit system operates on rules, that define what records will be captured in the audit logs. Audit rules are defined in the /etc/audit/audit.rules file or in the /etc/audit/rules.d/ directory. One rule is defined on a single line, as it was arguments to the auditctl utility.

As the events generated are dependent on the audit rules configured, it is recommended to first review the rules defined on the system before analyzing the audit logs. By default, no audit rules are configured, apart from eventual control rules often specific to the Linux distribution used. If no rules are configured, other components (pam, sshd, etc.) may still be sending events to the auditd daemon.

There are three types of audit rules:

  • Control rules: to modify the Linux Audit system behavior and its configuration.

  • File or watch rules: to audit access to files or directories.

    The watch rule format is: -w <FILE | FOLDER> -p <PERMISSION> -k <KEYNAME>.

    The access type to be audited, specified using the -p option, can be:

    • r: read of the file.

    • w: write to the file.

    • x: execute the file.

    • a: change in the file’s attribute.

    Example:

    # Generates events on modification to the "sudoers" file or files under the "sudoers.d" directory.
    -w /etc/sudoers -p wa -k sudoers_modification
    -w /etc/sudoers.d/ -p wa -k sudoers_modification
    
  • System call rules: to audit system calls (by any or the specified program).

    The system call rule format is: -a <ACTION>,<LIST> -S <SYSCALL> -F <FILTER_FIELD>=<FILTER_VALUE> [-F <FILTER_FIELD>=<FILTER_VALUE>] -k <KEY_NAME>

    Example:

    # Generates events on successful "connect" syscall from "/bin/bash".
    -a always,exit -F arch=b64 -F exe=/bin/bash -F success=1 -S connect -k "remote_shell"
    

The Neo23x0 auditd rules set can be used as a baseline, which may require fine tuning depending on the environnement and applications running on each system.

Information of interest

auditd events versus records

An auditd event can be split in multiple records. All records of the same event share the same timestamp (in the epoch format) and same unique identifier. Each record is associated with a given type. Records can be sometimes separated by hundreds other unrelated records.

The records are formatted as series of FIELD = VALUE pairs separated by a space. The available fields for auditd records, and their description, can be found in the Linux Audit Documentation.

Multiple events can be generated for a single operation, with each event potentially comporting multiple records. For example, the cat /etc/passwd command entered by a user in a TTY shell may generate:

  • An event for the execution of cat.

    Example rule: -w /usr/bin/cat -p x -k cat_exec.

  • An event for the access to /etc/passwd.

    Example rule: -w /etc/passwd -p warx -k etc_passwd

    This event would be split in the following records:

    type=SYSCALL msg=audit(1364481363.243:24287): arch=c000003e syscall=2 success=no exit=-13 a0=7fffd19c5592 a1=0 a2=7fffd19c4b50 a3=a items=1 ppid=2686 pid=3538 auid=1000 uid=1000 gid=1000 euid=1000 suid=1000 fsuid=1000 egid=1000 sgid=1000 fsgid=1000 tty=pts0 ses=1 comm="cat" exe="/bin/cat" subj=unconfined_u:unconfined_r:unconfined_t:s0-s0:c0.c1023 key="etc_passwd"
    type=CWD msg=audit(1364481363.243:24287):  cwd="/home/"
    type=PATH msg=audit(1364481363.243:24287): item=0 name="/etc/passwd" inode=409248 dev=fd:00 mode=0100600 ouid=0 ogid=0 rdev=00:00 obj=system_u:object_r:etc_t:s0  objtype=NORMAL cap_fp=none cap_fi=none cap_fe=0 cap_fver=0
    type=PROCTITLE msg=audit(1364481363.243:24287) : proctitle=636174002F6574632F706173737764
    
  • An event for a command being entered in a TTY shell.

Record types

The type field contains the type of the record:

Operation Field type(s)
User authentication and access USER_AUTH
USER_LOGIN
USER_LOGIN_SUCCESS
USER_LOGIN_FAILED
USER_AUTH_SUCCESS
USER_AUTH_FAILED
USER_START_SUCCESS
USER_START_FAILED
SESSION_TERMINATED
Process execution EXECVE
SYSCALL
Filesystem access PATH
For relative or absolute file access.

CWD
Current working directory, useful to reconstruct full path if a relative path has been recorded in PATH records.

OPENAT
Commands entered in a TTY console TTY
Commands entered by users USER_CMD
Full command-line of process PROCTITLE
The associated proctitle field MAY be encoded in hexadecimal.
Network socket connections SOCKADDR
The associated saddr field contains IP and port information, and can be interpreted directly at event generation (if log_format = ENRICHED is set), with ausearch -i, or simple scripting.
Account and group activity ADD_USER
ADD_GROUP

More record types are listed in the RedHat documentation.

Notable fields

  • The user and group identifiers (uid / gid, euid / egid, ruid / rgid, suid / sgid, and fsuid / fsgidof) of the process.

  • auid: audit identifier that, if present, identify the initial uid of the user (from the initial process). The auid remains the same even if the user’s identity changes (for instance through sudo or su, as the utilities start a new process with updated uid, euid, ruid, and suid).

    An auid not equal to the event’s uid / ruid can thus be an indication of privilege escalation through sudo or su.

  • tty and ses: respectively the terminal and session from which the audited process was invoked.

  • Some fields, such as the PROCTITLE command (proctitle field), may be stored in hexadecimal format. Keyword search on parameter values can thus be more complex and require adapted tooling.

  • For SYSCALL records, the aX field(s) define the arguments / parameters of the syscall, represented by unsigned long long integers and as such cannot be used to determine the values taken by the arguments.

Tool(s)

aureport and ausearch utilities

The aureport and ausearch utilities are packaged with the Linux Audit framework and can be used to analyze the auditd log files. aureport produces a synthesis of events, while ausearch can be used to search and filter the events.

# -i / --interpret: decodes some non-human readable values, such as epoch timestamps and hex encoded PROCTITLE commands, to plain-text.
# However, the uid are resolved based on the current system local accounts, potentially resulting in incoherent usernames.

# The -ts / --start and -te / --end: can be used to filter events based on their timestamp.
# The timestamp format depends on the local LC_TIME environment variable. For en_US.utf8 locale, the timestamp format is: MM/DD/YYYY hh:mm:ss.

# aureport.

# Displays the time range covered by each input file.
aureport -if <INPUT_FILE | INPUT_DIRECTORY> -t

# Generates a global synthesis: number of logins / authentications, failed logins / authentications, process and commands executed, tty keystrokes entered, etc.
aureport -i [-ts <START_TIME>] [-te <END_TIME>] -if <INPUT_FILE | INPUT_DIRECTORY>

# Generates a report on user logins (type=USER_LOGIN) with -l or authentications (type=USER_AUTH) with -au.
aureport -i [-ts <START_TIME>] [-te <END_TIME>] -if <INPUT_FILE | INPUT_DIRECTORY> -l
aureport -i [-ts <START_TIME>] [-te <END_TIME>] -if <INPUT_FILE | INPUT_DIRECTORY> -au

# Generates a report on the local accounts activity (creation or modification).
aureport -i [-ts <START_TIME>] [-te <END_TIME>] -if <INPUT_FILE | INPUT_DIRECTORY> -m

# Generates a report on program executions, which includes the session and auid of each process.
aureport -i [-ts <START_TIME>] [-te <END_TIME>] -if <INPUT_FILE | INPUT_DIRECTORY> -x

# Generates a report on TTY keystrokes (type=TTY) entered with TTY or commands (type=USER_CMD) with --comm.
aureport -i [-ts <START_TIME>] [-te <END_TIME>] -if <INPUT_FILE | INPUT_DIRECTORY> --tty
aureport -i [-ts <START_TIME>] [-te <END_TIME>] -if <INPUT_FILE | INPUT_DIRECTORY> --comm

# Generates a report on file access.
aureport -i [-ts <START_TIME>] [-te <END_TIME>] -if <INPUT_FILE | INPUT_DIRECTORY> -f

# ausearch.

# Filters on the specified event / message type (TTY, USER_CMD, EXECVE, PATH, USER_AUTH, etc.).
ausearch -i [-ts <START_TIME>] [-te <END_TIME>] -if <INPUT_FILE | INPUT_DIRECTORY>

# Filters on the specified user identifier (across uid, euid, and auid).
ausearch -i [-ts <START_TIME>] [-te <END_TIME>] -if <INPUT_FILE | INPUT_DIRECTORY> -ua "<UID>"

# Filters on the specified session or terminal identifier.
ausearch -i [-ts <START_TIME>] [-te <END_TIME>] -if <INPUT_FILE | INPUT_DIRECTORY> --session <SESSION_ID>
ausearch -i [-ts <START_TIME>] [-te <END_TIME>] -if <INPUT_FILE | INPUT_DIRECTORY> --terminal <TERMINAL_ID>

# Filters on the specified process identifier (PID) or parent process identifier (PPID).
ausearch -i [-ts <START_TIME>] [-te <END_TIME>] -if <INPUT_FILE | INPUT_DIRECTORY> -p <PID>
ausearch -i [-ts <START_TIME>] [-te <END_TIME>] -if <INPUT_FILE | INPUT_DIRECTORY> -pp <PID>

dissect target-query

The target-query tool, of the dissect Python framework, can be used to parse audit logs in CSV or JSON outputs. The fields will be partially decoded, as hex-encoded PROCTITLE commands are for instance not decoded to ascii.

# If processing an input folder directly, the folder may need to be placed in a tar archive.

target-query -f audit <TARGET> | rdump <--csv | --json | --jsonlines>

Sigma-based analysis with Zircolite and ChopChopGo

The Zircolite Python script (or standalone compiled binaries) and / or the ChopChopGo Go tool can be used to parse and process audit logs using Sigma rules, to generate a detection timeline of notable and potentially suspicious activity.

It should be noted however that Sigma rules for Linux are nowhere near the comprehensiveness level of Sigma rules for Windows, and only support basic detection.

# Prints the summary of the detections and outputs detailed results to a JSON file (by default).
python3 zircolite.py --auditd --ruleset rules/rules_linux.json [--csv] -e <AUDIT_LOG_FILE | AUDIT_LOG_FOLDER>

# Prints the detections in an ascii array by default, output format can be changed to JSON or CSV.
ChopChopGo -target auditd -rules ./rules/linux/auditd/ [-out <csv | json>] -file <AUDIT_LOG_FILE>

References



View on GitHub