Logging, Hardening & the Linux Security Baseline

Chapter 10 closes out the Linux track of the Foundations phase. It pulls together everything from the earlier chapters — permissions, users, processes, services, and networking — into the discipline of actually securing and monitoring a box: where logs live and how to read them, how auditd provides fine-grained visibility, and the concrete hardening steps that turn a default Linux install into something resembling a real security baseline. Every offensive technique explored in the rest of this notebook has a defensive mirror here; this chapter is that mirror.

---

Why Logging and Hardening Close Out This Track

Up to this point the track has been mostly about capability — how to navigate, script, filter text, manage processes, and move network traffic. This chapter is about accountability and reduction of attack surface: knowing what happened on a box after the fact, and reducing how much can happen on it in the first place. These two concerns — visibility and hardening — are the daily bread of SOC analysts, DFIR responders, and anyone doing compliance work (CIS benchmarks, PCI-DSS, SOC 2), and they are exactly what a penetration tester or red-teamer is trying to evade or bypass. Understanding both sides makes you dramatically better at each.

flowchart TD
    A[Linux Security Baseline] --> B[Logging & Visibility]
    A --> C[Hardening & Attack Surface Reduction]
    B --> B1[syslog / journald]
    B --> B2[auditd]
    B --> B3[Centralized log shipping]
    C --> C1[Minimal services / firewall]
    C --> C2[Access control - sudo, PAM, SSH]
    C --> C3[Mandatory Access Control - AppArmor/SELinux]
    C --> C4[CIS Benchmarks]

Foundations: Where Linux Logs Actually Live

The classic syslog-style logs

Path	Contents
/var/log/auth.log (Debian/Ubuntu) or /var/log/secure (RHEL/CentOS)	authentication events — logins, sudo usage, SSH connections
/var/log/syslog (Debian/Ubuntu) or /var/log/messages (RHEL/CentOS)	general system messages
/var/log/kern.log	kernel messages
/var/log/dmesg	boot-time kernel ring buffer
/var/log/apache2/ or /var/log/httpd/, /var/log/nginx/	web server access/error logs
/var/log/faillog, /var/log/lastlog	failed and successful login history, binary format
/var/log/wtmp, /var/log/btmp	login records (last, lastb read these)

tail -f /var/log/auth.log            # live-follow authentication events
grep "Failed password" /var/log/auth.log | tail -20
last -a                                 # recent logins with source host
lastb                                    # recent FAILED logins (btmp)
lastlog                                  # last login time per user

journald — the systemd-centric log store

Covered briefly in the previous chapter, journalctl is now the primary log interface on most modern distros, often in addition to traditional syslog files:

journalctl -xe                     # recent logs with extra explanation, end of output
journalctl -u sshd --since today   # one service, today only
journalctl -p err..alert            # priority range filter
journalctl --vacuum-time=7d          # prune logs older than 7 days

Why it works: journald stores logs in a structured binary format (not plain text), which is why you query it with journalctl instead of greping a file directly — this also means metadata (PID, UID, boot ID, systemd unit) is preserved per entry in a way flat text logs lose, which is exactly why forensic tooling prefers it when available.

auditd: Kernel-Level Visibility

auditd is the Linux Audit daemon — it hooks directly into the kernel’s audit subsystem to log security-relevant events (execve, file access, permission changes) with a level of detail syslog was never designed to provide.

sequenceDiagram
    participant Kernel
    participant AuditSub as Kernel Audit Subsystem
    participant Auditd as auditd daemon
    participant Log as /var/log/audit/audit.log
    Kernel->>AuditSub: syscall event (matches a loaded rule)
    AuditSub->>Auditd: emit audit record
    Auditd->>Log: write structured event

Installing and controlling auditd

sudo apt install auditd audispd-plugins   # Debian/Ubuntu
sudo systemctl enable --now auditd
sudo auditctl -l                            # list currently loaded rules

Writing audit rules

# Watch for ANY execution (a process being run) system-wide
sudo auditctl -a always,exit -F arch=b64 -S execve -k exec_tracking

# Watch a specific sensitive file for any write access
sudo auditctl -w /etc/shadow -p wa -k shadow_watch

# Watch for any changes to sudoers
sudo auditctl -w /etc/sudoers -p wa -k sudoers_watch

-k attaches a searchable “key” tag to matching events, which makes querying far easier:

sudo ausearch -k shadow_watch              # find all events tagged with this key
sudo ausearch -k exec_tracking -i          # -i "interprets" numeric UIDs/syscalls into names
sudo aureport --auth                        # summarized authentication report
sudo aureport -x                            # summarized executable report

Persisting rules across reboots means writing them into /etc/audit/rules.d/audit.rules rather than only running auditctl interactively (which is lost on reboot).

Why this matters: this is precisely the mechanism behind “we can see exactly what command an attacker ran, as what user, at what time” in almost every real Linux incident response writeup — auditd’s execve tracking is what turns “a process ran” into “here is the full command line, working directory, and parent process for every single execution on this box.”

Hardening: Reducing What Can Go Wrong

The CIS Benchmark mental model

The Center for Internet Security (CIS) publishes distro-specific hardening benchmarks (e.g., “CIS Ubuntu 22.04 Benchmark”) that are the industry-standard checklist for “what does a hardened Linux box look like.” You won’t memorize every control, but understanding the categories gives you the right mental model:

mindmap
  root((CIS-style Hardening Categories))
    Filesystem
      Separate partitions
      Disable unused filesystems
      Mount options - noexec/nosuid
    Services
      Disable unnecessary daemons
      Remove unused packages
    Network
      Firewall default-deny
      Disable IPv6 if unused
      Kernel network hardening sysctl
    Access control
      SSH hardening
      sudo least privilege
      Password policy / PAM
    Logging & auditing
      auditd rules
      Centralized log shipping
      Log rotation and retention

Concrete, immediately-actionable hardening steps

Firewall — default deny with ufw or nftables:

sudo ufw default deny incoming
sudo ufw default allow outgoing
sudo ufw allow 22/tcp     # only what you actually need
sudo ufw enable
sudo ufw status verbose

SSH hardening (revisit /etc/ssh/sshd_config):

PermitRootLogin no
PasswordAuthentication no
PubkeyAuthentication yes
AllowTcpForwarding no      # unless the box legitimately needs tunneling (see previous chapter)
MaxAuthTries 3

Kernel network hardening via sysctl:

/etc/sysctl.d/99-hardening.conf

net.ipv4.conf.all.rp_filter = 1        # anti-spoofing (reverse path filtering)
net.ipv4.icmp_echo_ignore_broadcasts = 1
net.ipv4.tcp_syncookies = 1             # SYN flood mitigation
net.ipv4.conf.all.accept_redirects = 0
kernel.dmesg_restrict = 1                # unprivileged users can't read kernel logs
sudo sysctl --system

Password and account policy via PAM (/etc/pam.d/common-password on Debian):

password requisite pam_pwquality.so retry=3 minlen=14 difok=3

fail2ban — automated brute-force banning:

sudo apt install fail2ban
sudo systemctl enable --now fail2ban
sudo fail2ban-client status sshd     # see currently banned IPs for the SSH jail

fail2ban watches log files (often auth.log/journald) for repeated failure patterns and dynamically inserts firewall rules to ban the offending IP for a configurable window — a lightweight, extremely common first line of defense against SSH brute forcing.

Mandatory Access Control: AppArmor and SELinux

Standard Linux permissions (covered in Chapter 3) are discretionary — the file owner decides who can access it. Mandatory Access Control (MAC) systems add a second, kernel-enforced layer that even root cannot casually bypass, confining what a process can do regardless of file permissions.

• AppArmor (Ubuntu/Debian default) — profiles are path-based and relatively easy to read: /etc/apparmor.d/usr.sbin.nginx restricts exactly what the nginx binary can touch, network-wise and filesystem-wise.
• SELinux (RHEL/CentOS/Fedora default) — label-based, more powerful and more complex, using “contexts” (system_u:object_r:httpd_sys_content_t) rather than plain paths.

# AppArmor
sudo aa-status                       # see loaded profiles and their mode (enforce/complain)
sudo aa-enforce /etc/apparmor.d/usr.sbin.nginx
sudo aa-complain /etc/apparmor.d/usr.sbin.nginx   # log-only mode, useful while tuning a new profile

# SELinux
getenforce                            # Enforcing / Permissive / Disabled
sudo setenforce 1                     # force Enforcing mode
ls -Z /var/www/html/                  # view SELinux contexts on files
sudo semanage fcontext -a -t httpd_sys_content_t "/srv/web(/.*)?"

Why this matters offensively too: a huge number of real-world web app exploits that “should” have led to full RCE are actually contained by SELinux/AppArmor — the process gets code execution, but the MAC policy blocks it from reading /etc/shadow or making outbound connections. Recognizing Permission denied errors that don’t match normal Unix permissions is a strong signal you’ve hit a MAC policy, not a Unix permission wall — checking dmesg | grep -i apparmor or ausearch -m avc confirms it instantly.

Building a Practical Hardening + Audit Lab

In a lab VM:

1. Install auditd and add a watch rule on /etc/passwd and /etc/shadow (-w /etc/shadow -p wa -k shadow_watch).
2. Trigger it: sudo cat /etc/shadow > /dev/null, then confirm the event with ausearch -k shadow_watch -i.
3. Set up ufw with a default-deny incoming policy, allow only SSH, and verify with nmap from another machine that no other ports respond.
4. Install fail2ban, then from a second machine deliberately fail SSH login 5+ times and confirm the source IP gets banned (fail2ban-client status sshd).
5. Check aa-status (or getenforce on an RHEL-family box) and note which profiles/policies are active versus complain/permissive — a real hardening review would flag every “complain”/“permissive” entry as a gap to close.

This lab exercises the full loop: harden something, generate the exact event that should be caught, and confirm your logging/audit layer actually caught it — the habit that separates “we have a firewall” from “we have evidence our firewall does anything.”

Advanced Techniques & Edge Cases

• Log tampering is a real attacker technique. Attackers with root frequently clear or edit /var/log/auth.log, wtmp, or shell history to hide tracks. Immutable, remote log shipping (forwarding logs to a separate syslog/SIEM server the moment they’re generated) defeats this — if the log left the box before it was deleted locally, the attacker’s cleanup is too late.
• Auditd rule ordering and performance. Overly broad rules (watching every execve on a busy production server) generate enormous log volume and can measurably affect performance; scope rules to specific paths/UIDs where possible, or accept the overhead deliberately for high-value hosts.
• noexec/nosuid mount options on /tmp and other user-writable partitions block a large class of “drop a binary in /tmp and execute it” techniques used by both attackers and CTF privesc chains — but be aware some legitimate software (and some CTF exercises specifically) expect /tmp to be executable, so test before deploying broadly.
• Rotation and retention tuning (logrotate, journalctl --vacuum-size) matters for compliance: many standards (PCI-DSS) mandate a minimum retention period, and a box that silently rotates/deletes logs too aggressively can fail an audit even if the logging itself was configured correctly.

Real-World Application

• Compliance-driven hardening: CIS Benchmarks, PCI-DSS, HIPAA, and SOC 2 audits all reference concrete Linux configuration checks nearly identical to what’s in this chapter — hardening scripts built around CIS controls are standard deliverables in GRC and cloud security engagements.
• DFIR case studies: real breach post-mortems (e.g., numerous publicly documented ransomware and cryptomining incidents) repeatedly cite “no centralized logging,” “auth.log rotated/deleted before analysis,” or “auditd not installed” as the reason root cause could not be conclusively established — this chapter’s content is literally the difference between a resolvable incident and an unresolved one.
• Pentest reporting: a professional pentest report routinely includes a “detective controls” or “logging & monitoring” finding alongside pure vulnerability findings — testers are expected to comment on whether their own activity would have been detected, which requires understanding exactly what’s covered in this chapter.

Detection & Defense: Blue Team Angle

This entire chapter is the blue team angle, but concretely:

• Ship logs off-box immediately (syslog-ng/rsyslog forwarding, or an agent like Filebeat/Wazuh) so local tampering can’t erase evidence.
• Alert on auditd’s own health — auditd stopping unexpectedly, or auditctl -l returning fewer rules than expected, is itself a signal worth alerting on (it’s a known attacker technique to disable audit logging before further action).
• Baseline aa-status/getenforce across your fleet — any host silently running in permissive/complain mode has quietly lost a real layer of defense, often without anyone noticing.
• Watch /etc/sudoers, /etc/passwd, /etc/shadow, and SSH config for unauthorized changes via file integrity monitoring plus the auditd watches shown above — these four files/directories cover a disproportionate share of real persistence and privilege escalation activity.

Blue Team CTF / detection challenge angle — A very common exercise: given an audit.log/auth.log excerpt, identify what an attacker did, in order — typically login, privilege escalation attempt via sudo or a SUID binary, a suspicious execve of an unusual binary path, then log tampering. Reconstructing that timeline purely from ausearch/grep output is a recurring DFIR-flavored CTF category worth deliberate practice.

Bug Bounty Angle — Logging/hardening gaps aren’t directly reportable bugs in most bug bounty programs (they’re infrastructure config, usually out of scope), but they matter for two reasons: (1) some programs specifically include “insufficient logging” as a valid finding under OWASP’s broader vulnerability categories when chained with another bug, and (2) demonstrating that a vulnerability you found would have gone completely undetected (no logging, no alerting) is powerful supporting evidence when arguing for higher severity in your writeup.

CTF Angle — Forensics categories frequently hand you a log file (auth.log, audit.log, or a full disk image) and ask “what did the attacker do and in what order.” Tools: ausearch, aureport, grep/awk (Chapter 7), and last/lastb. Worked pattern: start with lastb for failed logins to find the brute-force window, grep "Accepted" in auth.log for the successful login, then ausearch -ts recent -k exec_tracking (if auditd was running) to reconstruct exactly what commands ran afterward.

Common Mistakes & How to Overcome Them

Mistake	Symptom	Fix
Assuming default install is “secure enough”	Unnecessary services/ports exposed	Explicitly review and apply a CIS-style checklist
Local-only logging	Attacker deletes/edits logs, no evidence remains	Ship logs off-box in real time
Overly broad auditd rules	Massive log volume, missed signal in noise	Scope rules to specific sensitive paths/syscalls
Treating AppArmor/SELinux denials as “broken app” and disabling MAC entirely	Loses a real defense layer	Use complain/permissive mode to tune the profile instead of disabling it
No automated brute-force protection	SSH gets hammered continuously	Deploy fail2ban (or equivalent) as a baseline, not an afterthought

Final Revision / Summary

• Linux logs live in two overlapping worlds: classic flat-text syslog files (/var/log/auth.log, /var/log/syslog) and the structured journald store queried via journalctl.
• auditd hooks the kernel directly, giving execve-level and file-access-level visibility that syslog alone can’t provide — auditctl to set rules, ausearch/aureport to query them.
• Hardening follows a repeatable shape: minimize attack surface (firewall default-deny, disable unused services), harden access control (SSH config, sudo, PAM password policy), add automated response (fail2ban), and add a mandatory access control layer (AppArmor/SELinux) as defense in depth.
• CIS Benchmarks are the industry reference checklist for “what does hardened actually mean” per distro.
• Mnemonic: “See it, Stop it, Seal it” — logging/auditd (see it), firewall/fail2ban/SSH hardening (stop it), AppArmor/SELinux (seal it even if the first two fail).

Cheat Sheet / Quick Reference

# Logs
tail -f /var/log/auth.log
journalctl -u sshd --since today
last -a / lastb / lastlog

# auditd
sudo auditctl -w /etc/shadow -p wa -k shadow_watch
sudo ausearch -k shadow_watch -i
sudo aureport --auth

# Firewall
sudo ufw default deny incoming
sudo ufw allow 22/tcp
sudo ufw enable

# fail2ban
sudo fail2ban-client status sshd

# MAC
sudo aa-status            # AppArmor
getenforce                 # SELinux

# sysctl hardening
sudo sysctl --system

Practice Labs & Resources

• CIS Benchmarks (cisecurity.org) — download the free PDF for Ubuntu or your distro of choice and work through the “Level 1” controls as a checklist against a lab VM.
• TryHackMe “Linux Hardening” and “Auditd” rooms — guided, graded exercises building auditd rules and confirming they fire.
• Lynis (github.com/CISOfy/lynis) — a free automated hardening-audit tool; run it against your lab VM before and after applying this chapter’s steps to see a measurable score improvement.
• HackTheBox / TryHackMe “DFIR”-tagged rooms — practice reconstructing an attacker timeline purely from log excerpts, mirroring the CTF Angle worked pattern above.
• OpenSCAP / oscap — open-source tooling that can automatically evaluate a box against a CIS or DISA STIG profile, useful for turning this chapter’s manual checklist into an automated, repeatable scan.