Phishing & social engineering

Phishing

A social-engineering attack that impersonates a trusted entity to trick a person into surrendering credentials, money, or access — by email, SMS, voice, QR code, or OAuth consent.

Phishing is a social-engineering attack in which the attacker impersonates a trusted entity — a bank, a colleague, a SaaS vendor, a courier — to manipulate a target into handing over credentials, money, or access to an internal system. It is the single most common initial-access vector in security incidents: the Verizon DBIR has identified phishing and pretexting as the dominant human-element vector for over a decade.

The format has evolved well beyond the bulk email of the 2000s. Defining properties of modern phishing:

  • Multi-channel. Email remains dominant, but vishing (voice), smishing (SMS), qishing (QR codes), and OAuth consent phishing now make up a meaningful share of incidents. ANSSI’s annual Panorama de la cybermenace tracks the channel mix in France.
  • Targeted. Generic “Nigerian prince” lures are residual noise. The damaging variants are spear-phishing against specific roles (finance, IT, execs) and BEC against payment-approval chains.
  • MFA-aware. Attacker kits like EvilProxy and Tycoon proxy MFA challenges in real time; older “phishing-resistant” claims no longer hold for push and TOTP. Only FIDO2 / passkeys are structurally resistant.
  • AI-assisted. Generative AI removes the language-quality tell on which legacy filters and trained users relied. A 2024 sentence-by-sentence French-language lure is now indistinguishable from a legitimate one to most readers.

The buyer-side mistake is to treat phishing as a filter problem. Secure email gateways and AI detectors catch the bulk; the residual that lands in inboxes is, by definition, the slice that defeated detection. What matters from that point is human behavior: report rates, time-to-report, repeat-clicker concentration. That is the surface on which a behavior-centered program — see behavioral KPI — is measured.

Related terms

See also