Topic 1 ยท Deep Dive

1.4 Human-Centered Security Habits

Verify the Envelope, Not the Display Name

The "From" line is free text; machines should be consulted instead. SPF (RFC 7208) publishes which IPs may send for a domain, e.g. v=spf1 include:_spf.google.com ~all. DKIM (RFC 6376) attaches an RSA or Ed25519 signature over selected headers and the body, with the public key in a DNS TXT record. DMARC then requires the visible From to align with the SPF or DKIM domain and can publish p=reject to kill forgeries outright. In a client, open the raw headers and read Authentication-Results: a spf=pass on a totally different envelope-from domain is a friendly-from spoof. Also scan for homoglyph registrations rendered as punycode, such as xn--paypa-btf8k.com.

Read URLs Like a Parser

A URL is scheme://user:pass@host:port/path?query#frag, and browsers paint only part of it. The host is the last label before the first slash, so paypal.com@evil.example, under any scheme, sends you to evil.example, not PayPal. Percent-encoding hides structure: %2e%2e is .., and a trailing .paypal.com.evil.example is still evil. Compare the registrable domain against the Public Suffix List rather than eyeballing. Paste the link into a plain text field to reveal embedded newlines or padding, and remember a QR code encodes the identical string — it is not safer, just un-inspectable.

Break the Urgency Loop Out-of-Band

Every manipulation shares one mechanism: compressed time plus a channel the attacker chose. Counter it by moving to a channel you chose. If an email claims a fraud alert, call the number printed on your last statement or use the official app — never the number or link in the message. For a payment change, the control is a callback policy: new bank details are activated only after a call to the vendor number already on file. For voice calls, agree on a challenge phrase or safe word that a cloned voice cannot know, and ask something with no searchable answer.

Report First, Then Delete

Deleting a lure destroys the evidence your filters need to learn. Forward it intact (headers included) to your IT report button or to reportphishing@apwg.org, then delete. On the account side, habits beat heroics: a password manager, passkeys instead of SMS where offered, and auto-updates left enabled — most successful phishing works on stale software, not on clever writing.

Architecture Diagram

Human decision chain for inbound messages Incoming message Verify sender SPF, DKIM, DMARC align Inspect link registrable domain wins Safe: proceed Report and delete
Two mechanical gates — authenticate the sender, then read the real host — route every message to either proceed or report.

Key Takeaways

« Back to Topic 1« 1.3 / 1.5 »