Phishing Response Playbook
One reported email becomes a campaign. A campaign becomes a credential harvest. A credential harvest becomes a ransomware precursor. The playbook that handles a single phishing report is not the playbook that handles what it might actually be.
A user forwards a suspicious email to the security team at 9:15 AM on a Tuesday. The email claims to be from DocuSign, asks the user to review a document, and contains a link to a convincing login page. Standard phishing. The analyst opens a ticket, pulls the email headers, detonates the URL in a sandbox, blacklists the domain, and marks the ticket resolved by 9:40 AM.
At 2 PM, another analyst notices something in the proxy logs: 23 distinct internal hosts contacted the same domain between 8 AM and 9 AM the previous day — before the first report came in. A quick SIEM search shows the same phishing email was delivered to 1,400 users over a 3-day campaign. 23 users clicked the link. 7 of those users typed credentials into the phishing page, which was confirmed by reviewing the POST requests in proxy logs.
The 25-minute "resolved" ticket was not a phishing response. It was a single-email cleanup that left a credential theft campaign completely unaddressed. This chapter is about the difference between those two things — and how to run the response that actually matters.
Single Email vs. Campaign: The First Decision
The first action in any phishing response is to determine whether you are dealing with a single targeted email or a campaign. These require fundamentally different responses — the same way that finding one infected host versus finding a ransomware deployment affecting 40 hosts requires different responses. The trigger to escalate from "single email" to "campaign response" is finding more than one recipient of the same or similar message.
Phishing Response Scope Decision Tree
══════════════════════════════════════════════════════════════════════
User reports suspicious email
│
▼
Search email gateway for similar messages
(same sender domain, same subject pattern,
same URL domain, same attachment hash)
│
├── Found: 1 recipient (the reporter only)
│ → Single-email response
│ Analyze email, check URL, block sender/domain,
│ notify user, close ticket
│
└── Found: Multiple recipients (or: sender domain
used across multiple messages)
→ CAMPAIGN RESPONSE
Scale changes everything:
• How many users received it?
• How many clicked? (proxy logs, URL-click tracking)
• How many submitted credentials?
• Over what time period?
These answers determine the IR scope
Phishing campaigns are not sent one email at a time. Threat actors run bulk sending infrastructure that delivers thousands of messages in automated waves. By the time one user reports a phishing email, the campaign has usually been running for hours or days. The first reporter is almost never the first recipient — they are simply the first person who noticed and bothered to report it. Responding to the single reported email while ignoring the campaign is the equivalent of treating one symptom while the disease spreads. The correct default assumption for any phishing report is: this is a campaign until proven otherwise. Proving otherwise requires checking the email gateway — not assuming.
Email Header Analysis: Reading the Mail Trail
Email headers are the chain of custody for an email message. Every mail server that handled the message adds a Received header. Reading them correctly tells you where the message actually came from — not where it claims to have come from.
Reading Received Headers — Bottom to Top
══════════════════════════════════════════════════════════════════════
Email headers are added in reverse order — the FIRST Received header
at the bottom of the chain is where the email originated.
The LAST Received header at the top is your own mail server.
Received: from mail.yourcompany.com (your MX) ← YOUR SERVER (top)
Received: from smtp-relay.sendgrid.net ← intermediate relay
Received: from outbound.evil-domain.net ← claimed sending server
Received: from [192.168.1.100] ← ACTUAL origin IP (bottom)
Key fields to examine:
──────────────────────────────────────────────────────────────────
From: Display name and address in the email client.
Trivially spoofed. Never trust without verification.
Return-Path: Where bounce messages go. Different from From:
in spoofed email. Compare these.
X-Originating-IP: Many services include the sender's real IP here.
Check this against threat intel.
Authentication-Results: Your mail server's SPF/DKIM/DMARC results.
This is where the authentication truth lives.
Message-ID: Unique identifier. Useful for searching across
gateway logs to find all copies of the same message.
SPF, DKIM, and DMARC — What the Authentication Results Mean
Modern email authentication uses three layered mechanisms. Understanding what each checks and what a failure means tells you how sophisticated the phishing attempt was.
| Mechanism | What it checks | PASS means | FAIL means |
|---|---|---|---|
| SPF | Whether the sending IP is authorized to send mail for the claimed domain | The sending IP is listed in the domain's SPF record | The sending IP is NOT authorized — likely spoofed sender domain or unauthorized relay |
| DKIM | Cryptographic signature on the email body and headers, verified against a public key in DNS | The email was cryptographically signed by the claimed sender's mail server | The signature is missing or invalid — email may be tampered or from an unauthorized sender |
| DMARC | Whether SPF or DKIM aligns with the From: domain, and what to do on failure | Email passes the domain owner's authentication policy | The email doesn't align with the domain's authentication requirements — check DMARC policy (none/quarantine/reject) |
SPF and DKIM each answer independent questions. DMARC answers the policy question: given that SPF and/or DKIM passed or failed, what should the receiving mail server do? A DMARC policy of "reject" means the receiving server should refuse to deliver email that fails SPF and DKIM alignment. A policy of "none" means the domain owner is still collecting data and not enforcing — meaning failing email is still delivered. Many legitimate organizations publish DMARC at p=none, which means their domain name can be used in phishing emails with no technical barrier to delivery. The authentication-results header in a phishing email tells you whether the attacker spoofed the domain outright (SPF fail) or registered a lookalike domain that passes all authentication (SPF pass, DKIM pass, DMARC pass — because it's a legitimately registered domain sending legitimate email, just impersonating someone else).
URL Analysis: Never Click Directly
The URL in a phishing email is the primary weapon. Analyzing it correctly requires detonating it safely — in a sandbox environment that will not expose production credentials, will not be traced back to your organization's IP, and will capture the full behavior of the page and any redirects.
Phishing URL Analysis Workflow
══════════════════════════════════════════════════════════════════════
Received URL: https://login.microsoftonline-secure.net/auth/v2/token
Step 1: Check the domain registration
→ registrar.search(microsoftonline-secure.net)
→ Registered 3 days ago. Registrar: NameCheap. Privacy-protected.
→ Typosquatting indicator: "microsoftonline" in domain name
→ Real Microsoft domain: microsoftonline.com (different TLD suffix)
Step 2: Check threat intel before detonating
→ VirusTotal URL scan (does not click, uses cached analysis)
→ URLScan.io (passive scan)
→ Check URL against known phishing feeds (PhishTank, OpenPhish)
Step 3: Detonate in sandbox (if not already categorized as malicious)
→ Browserling, AnyRun (URL mode), Joe Sandbox, Hybrid Analysis
→ Capture: full redirect chain, final page content, form fields,
POST destination, any JavaScript behavior, cookies set
Step 4: Extract indicators from detonation
→ All domains in redirect chain → add to blocklist
→ POST destination (where credentials are sent) → may be different
from the initial URL (exfil domain)
→ Any file downloads triggered → hash and analyze separately
Sophisticated phishing pages detect the source IP of the visitor. Requests from known corporate IP ranges, security vendor IP ranges, or bot detection patterns may show a benign page — a legitimate-looking Microsoft login page with no malicious behavior — while delivering the actual credential-harvesting page to real victims. If you visit the phishing URL from your corporate IP and it looks clean, that does not mean it is clean. Always detonate in a third-party sandbox with a residential or cloud IP that has no association with your organization or security vendors. This is one of the more effective phishing kit defenses and it trips up analysts who rely on personal URL inspection.
Determining Campaign Scope
Once a campaign is confirmed, you need three numbers: how many people received the email, how many clicked, and how many submitted credentials. Each requires a different data source.
| Question | Primary data source | How to query | What to look for |
|---|---|---|---|
| How many received it? | Email gateway / MTA logs | Search by sender domain, subject line pattern, URL domain, attachment hash, or Message-ID range | Delivery receipts to internal addresses — count distinct recipients |
| How many clicked the URL? | Proxy / web gateway logs | Search for connections to the phishing domain(s) identified in URL analysis, within the campaign delivery window | HTTP/HTTPS connections to phishing domains — source IP maps to internal host, then to user via DHCP logs |
| Did URL-click tracking fire? | Email security platform (Proofpoint TAP, Mimecast, Defender for Office) | URL-click reports in the platform console — these platforms rewrite URLs and log every click | Click events with user identity and timestamp — most reliable source if available |
| Who submitted credentials? | Proxy logs (POST requests) | Among hosts that connected to the phishing domain, filter for HTTP POST requests — especially to credential submission endpoints | POST to /auth, /login, /submit, or similar paths — the POST body may contain credential data if proxy performs SSL inspection |
The Credential Compromise Assumption
This is the most important judgment call in phishing response: if a user clicked a link and landed on a page with a login form, treat their credentials as compromised regardless of what the user says they did. This is not a question of trust — it is a question of risk management under uncertainty.
Why User Self-Reporting on Credential Submission Is Unreliable
══════════════════════════════════════════════════════════════════════
User reports: Reality (from proxy logs):
"I didn't click it" → POST request to phishing domain
from their host at 08:43:22
"I clicked but didn't enter → POST body contains user@corp.com
anything" and a string matching password
pattern (if SSL inspection enabled)
"I entered my work email but → The phishing kit collects whatever
not my password" is typed in the email field;
password field not filled != not
attempted
"I entered a fake password → Phishing kits accept any input and
to test it" relay it; attacker got your "fake"
password; if it matches any real
password you use, you're compromised
RULE: Credential submission to a phishing page = credentials compromised.
Reset immediately. Do not wait for confirmation of attacker use.
The cost of resetting a credential that was not actually compromised is: the user spends 3 minutes setting a new password and may be mildly inconvenienced. The cost of not resetting a credential that was compromised is: the attacker uses that credential for account takeover, lateral movement, or as the initial access vector for a ransomware deployment. These costs are not comparable. The rule is: when in doubt, reset. The only time you should not reset based on a credential compromise suspicion is if resetting the credential would cause more damage than the potential compromise — which is rare and requires specific justification.
User Notification and Credential Reset at Scale
When 1,400 users received a phishing campaign and 23 clicked, you have two distinct notification tasks: broad notification to all recipients (lower urgency), and targeted notification to clickers with immediate action required (higher urgency).
Two-Tier Notification Approach
══════════════════════════════════════════════════════════════════════
TIER 1: All campaign recipients (1,400 users)
─────────────────────────────────────────────
Communication: IT/Security all-hands email or intranet post
Urgency: Medium — informational, no immediate action for most users
Message:
"We identified a phishing campaign that delivered emails
appearing to be from DocuSign between [dates]. If you
received this email, do not click any links. If you
believe you may have clicked a link, contact the help
desk immediately at [contact]."
Goal: Surface additional clickers who did not show in logs
TIER 2: Confirmed clickers (23 users)
──────────────────────────────────────
Communication: Direct phone call or in-person (not email —
if their email is compromised, they may not receive it safely)
Urgency: HIGH — immediate password reset required
Message:
"Your account credentials may have been compromised.
We need you to reset your password immediately and
verify your MFA enrollment. An IT security analyst
will walk you through this now."
Additional actions:
• Force immediate password reset
• Review MFA enrollment (attacker may have added device)
• Check for new inbox rules (sign of account compromise)
• Review recent sign-in activity for anomalies
When Phishing Is a Ransomware Precursor
Modern ransomware groups use phishing as their initial access vector. This means every significant phishing campaign has a non-trivial probability of being the opening move in a ransomware deployment. The question to answer in parallel with the credential response is: has any of the credential access already been used?
| Check | What you're looking for | Data source | What it means if found |
|---|---|---|---|
| New sign-ins from unusual IPs or geographies | Successful authentications using compromised credentials from IPs not previously seen for those accounts | Azure AD sign-in logs, Okta System Log, on-prem AD 4624 events | Credential is actively being used — escalate to full account takeover IR |
| New inbox rules | Rules forwarding email to external addresses, deleting security team notifications, marking security alerts as read | M365 Unified Audit Log (Set-InboxRule events), Exchange admin | BEC/account takeover — attacker is maintaining persistent access to email |
| New MFA devices enrolled | MFA devices added to accounts that were phished, especially in the hours after credential submission | Azure AD MFA enrollment events, Okta device enrollment log | Attacker has enrolled their own MFA device — do not just reset password, also remove unauthorized devices |
| Lateral movement from compromised accounts | Authentication events from compromised accounts to systems those accounts don't normally access | AD Event 4648, Kerberos ticket requests to unusual services | Active intrusion in progress — ransomware deployment timeline begins at successful lateral movement |
The 2023 MGM Resorts incident began with Scattered Spider social engineering the IT help desk — a variation of phishing where the attacker calls IT support, impersonates an employee, and convinces IT to reset MFA. Within hours of gaining that initial access, the group had moved laterally into VMware ESXi infrastructure and deployed ALPHV/BlackCat ransomware. The total disruption cost MGM an estimated $100 million. The initial "phishing" response — credential compromise by a single account — was the entire attack. Every credential compromise from a phishing campaign is potentially the opening move. The question is not "did someone fall for phishing" but "what did the attacker do after they got the credential."
Phishing Response Playbook Summary
| Time | Action | Owner |
|---|---|---|
| T+0:00 | Receive phishing report; begin email header analysis | L1 analyst |
| T+0:10 | Search email gateway for campaign scope (all recipients, delivery dates) | L1 analyst |
| T+0:15 | If campaign confirmed: escalate, search proxy logs for clickers | L1 → L2 |
| T+0:20 | Detonate URL in sandbox; extract all IOCs (domains, IPs, hashes) | L2 analyst |
| T+0:25 | Block phishing domains at email gateway, proxy, and DNS | L1 or platform team |
| T+0:30 | Identify credential submitters from proxy POST logs | L2 analyst |
| T+0:45 | Begin Tier 2 notifications (direct contact with credential submitters) | L2 + Help Desk |
| T+1:00 | Force password reset + MFA review for all confirmed credential submitters | Help Desk + IAM team |
| T+1:00 | Check for credential use (sign-in anomalies, new inbox rules, MFA enrollments) | L2 analyst |
| T+2:00 | Tier 1 notification to all campaign recipients | IR lead or comms |
| T+4:00 | If credential use detected: escalate to full account takeover IR | IR lead |
| T+24:00 | Close loop: confirm all credential resets complete, no remaining active sessions | L2 analyst |
Q & A
Q: Proxy logs show a user connected to the phishing domain but there are no POST requests logged. Does that mean they didn't submit credentials?
Not necessarily. The absence of POST requests in proxy logs has several possible explanations: (1) The proxy does not perform SSL inspection — if the phishing site used HTTPS and the proxy only decrypts specific categories or domains, the POST body (and even the POST method) may not be logged, only the TCP connection. (2) The phishing kit used a technique that avoids obvious POST requests — some phishing kits submit credentials via JavaScript XMLHttpRequest or fetch() calls that may be categorized differently by the proxy. (3) Log retention gap — if the incident was discovered days later, POST-level logs may have rolled over while connection logs are retained longer. (4) The user connected, saw a suspicious page, and navigated away before entering anything. Treat cases (1), (2), and (3) as credential compromise assumed. Only (4) — confirmed by also checking EDR browser history if available — justifies lower confidence. The safe call is: connected to phishing page = treat as credential compromised.
Q: How do you identify which user a proxy log IP address belongs to when you have DHCP-assigned IPs?
IP-to-user resolution requires correlating multiple data sources: (1) DHCP lease log — the DHCP server logs which machine received which IP at what time. Cross-reference the proxy connection timestamp with the DHCP lease for that IP. This gives you the hostname. (2) Authentication log — once you have the hostname, check AD authentication logs for logins to that machine around the relevant time. Event ID 4624 with logon type 2, 10, or 11 tells you who was logged in. (3) EDR — if the host has an EDR agent, it maps the hostname to a user continuously. The EDR console can tell you exactly who was logged in at 08:43 on Tuesday. If none of these yield a definitive answer (shared workstation, kiosk machine, VDI session), fall back to checking the email gateway: if the phishing email was delivered to that IP's user account at the right time, the correlation is strong even without definitive proof of who was at the keyboard.
Q: The phishing campaign used a lookalike domain with valid SPF, DKIM, and DMARC. How did it get through email filtering?
Lookalike domain phishing is specifically designed to pass email authentication checks. The attacker registers a domain like "microsoftonline-auth.com" — different from "microsoftonline.com" — and sets up valid SPF, DKIM, and DMARC records for that registered domain. All authentication checks pass because the email genuinely was sent from that domain. Email authentication verifies that the sender controls the sending domain — it does not verify that the domain is the legitimate organization. The defense against lookalike domains is not authentication — it is display name analysis, URL reputation analysis, and machine learning-based similarity detection. Enterprise email security platforms (Proofpoint, Mimecast, Defender for Office 365) have lookalike domain detection specifically for this. If you're being hit by lookalike campaigns, review whether your email security platform's impersonation detection is enabled and properly configured for your most commonly impersonated senders (Microsoft, DocuSign, Zoom, your own CEO's name).
Q: 7 users submitted credentials. You reset all 7 passwords. Two hours later, 2 of those accounts show new sign-in activity from a foreign IP. What happened?
The attacker likely had a session already established before the password was reset. Modern authentication systems use session tokens (cookies, OAuth tokens, JWTs) that remain valid even after a password reset — resetting a password invalidates the ability to log in with that password, but it does not automatically revoke existing sessions. If the attacker authenticated before the reset, they may have a valid session token that continues to work for hours or days. The fix: for accounts with confirmed credential compromise, password reset is necessary but not sufficient. You must also revoke all active sessions. In Azure AD/Entra ID, this is "Revoke Sign-In Sessions" on the user account, which invalidates all refresh tokens and forces re-authentication. In Okta, terminate all active sessions. In on-prem environments, invalidating Kerberos TGTs requires either waiting for the ticket lifetime to expire (typically 10 hours) or disabling the account temporarily. This is a common post-phishing gap: teams reset passwords, don't revoke sessions, and then wonder how the attacker is still active.
Q: You've blocked the phishing domain. Users are still reporting receiving the emails. Why?
Phishing campaigns use multiple domains and redirect chains. Blocking one domain stops traffic to that domain but not to the campaign infrastructure as a whole. Likely explanations: (1) The campaign is sending from multiple sender domains — you blocked one, but others are still delivering. Search the email gateway not just for the blocked domain but for the email template pattern (subject line, body text signature, attachment type). (2) The phishing link redirects through multiple domains — you blocked the final credential-harvesting domain but not the intermediate redirect chain. Users clicking links in email delivered before the block go through redirects to the new domain. (3) The attacker rotated domains when they detected the block — this is common in sophisticated campaigns. The response: expand your search to find all related domains (look for newly registered domains with similar naming patterns, check VirusTotal for related infrastructure, look at WHOIS for the same registrant or registrar used for the blocked domain), and block the campaign pattern rather than individual domains.
Q: A user submitted credentials but the phishing page appeared to be for a different company — not yours. Why would an attacker phish your users with a page impersonating another company?
Several scenarios: (1) Credential stuffing preparation — the attacker wants usernames and passwords that people reuse across services. If they harvest your users' emails and whatever password they type, they can test those credentials against other services (bank, email, LinkedIn). The "company" being impersonated is just the bait, not the target. (2) Third-party supplier compromise — the attacker is targeting a specific SaaS vendor your company uses. By harvesting credentials on a page impersonating that vendor, they get the credentials your users use for that service. (3) Business email compromise setup — the phishing page impersonates a vendor in your supply chain. The goal is to compromise a vendor employee account to send convincing BEC emails to your finance team, not to directly compromise your user accounts. The response is the same (credential reset, session revocation, check for use) but the threat model is different — the risk extends beyond your organization to your vendor relationships.