Incident Response Forensics

Collect and analyze evidence without destroying it.

1. Preserve Before You Poke

• Take memory capture first (before heavy processes): WinPMem, LiME, or vendor snapshot
• Disk image via write-blocker or verified cloud snapshot; hash everything (SHA-256), record chain of custody
• Collect network evidence: firewall/flow logs, PCAP if available, proxy and DNS logs
• Never investigate on the live system beyond read-only collection — attacker tooling notices

2. Volatile Data Order

1. Network connections, routing tables
2. Running processes and services, loaded modules
3. Logged-in sessions, clipboard (if interactive)
4. Command histories, scheduled tasks, autostart

3. Artifact Analysis

Windows

• Shimcache/Amcache, Prefetch, SRUM for execution history
• Event logs: 4624/4625 (logons), 4688 (process creation), 7045 (service install), PowerShell 4104
• Registry: Run keys, Services, USN journal for file deletion
• Lateral movement: RDP logs (4776/4624 type 3), scheduled tasks, WMI activity

Linux

• /var/log/auth.log, journalctl, .bash_history, wtmp/last
• Cron, systemd timers, LD_PRELOAD in /etc, immutable-bit anomalies
• Unauthorized SSH keys in authorized_keys, passwd modifications

Cloud

• CloudTrail/data plane logs, unusual AssumeRole chains, new access keys
• Control-plane vs data-plane timeline separation

4. Timeline Construction

• Normalize all artifacts into a super-timeline (log2timeline/PlutoScope-style)
• Anchor on the initial access indicator, then walk forward (persistence, spread, exfil) and backward (recon, delivery)
• Separate facts from inference; mark confidence levels

Output

• IOCs with context (hash, IP, domain, first/last seen)
• Attack narrative mapped to ATT&CK tactics with evidence citations per step
• Root-cause hypothesis with supporting and contradicting evidence