Evasion

Evasion audio micro-course (transcript)

Three short podcasts about “Evasion”. Listen with the interactive player.

Podcast 1 — Introduction & meaning

Narrator: Welcome. Today's word is evasion — spelled E-V-A-S-I-O-N, pronounced ih-VAY-zhun, IPA slash ɪ-ˈveɪ-ʒən slash.

Narrator: Evasion is a noun. It names the action of escaping or avoiding something — a duty, a question, a pursuer, or an obligation — especially through cleverness or deliberate manoeuvring.

Narrator: In plain terms, evasion is getting away from, or cleverly avoiding, something you are supposed to face.

Narrator: The word descends from the Latin evasio, built from evadere — to go out or escape — combining e-, out, with vadere, to go or walk. It entered English in the fifteenth century via Old French.

Narrator: Its earliest English use was quite physical — evasion described an actual escape from danger or captivity. A prisoner's evasion. A fox's evasion of the hounds.

Narrator: By the seventeenth century, the meaning had moved inward. You could evade a question with a slippery answer, evade a duty by bureaucratic delay, evade a moral obligation by looking the other way.

Narrator: Today, evasion lives in several very specific and very serious domains. Tax evasion is a criminal offence — the illegal non-payment of tax. Draft evasion. Truth evasion in public life. The evasion of a direct answer in a political interview.

Narrator: But it also appears in sport — a winger's evasion of a tackle — and in computing — evasion techniques used by malware to dodge detection software.

Narrator: The tone is neutral to mildly negative. Evasion almost always implies something that should have been faced was not.

Narrator: Remember this: evasion is avoidance dressed up to look like something else.

Speaker A: I asked my manager three times about the budget and got three completely different non-answers. That's evasion of the most frustrating kind.

Speaker B: Classic organisational evasion. The question is real, the answer is inconvenient, and so the conversation goes round in circles.

Speaker A: Exactly. And the word evasion captures something avoidance doesn't quite — the sense of active manoeuvring. It's not just avoiding, it's dodging with intent.

Speaker B: Right. Avoidance can be passive — I simply didn't go. Evasion is more active — I went, but I slipped through the net.

Speaker A: Good distinction. And there's a legal shading too. Tax avoidance is legal, if often aggressive. Tax evasion is criminal.

Speaker B: Exactly. The same difference — one works within the system's loopholes, the other sidesteps the system entirely. Evasion is the darker word.

Speaker A: And in conversation, we often call vague political answers evasions — not necessarily lies, but structured to prevent any commitment.

Speaker B: Right. A skilled political evasion leaves no fingerprints. The question was asked; an answer was produced; but nothing was actually said.

Speaker A: Close synonyms — avoidance, circumvention, dodging. Circumvention feels more legal and technical; dodging is physical and casual; evasion is the most deliberate and often the most morally loaded.

Speaker B: Common mistake — using evasion when you just mean avoidance. Not returning a friend's call isn't evasion; it's just avoidance. Evasion involves active manoeuvring under some kind of obligation.

Speaker A: A sharp word. It implies you knew what you were doing.

Speaker B: And usually, you did.

Speaker B: Today's word is evasion. In prompting, it's vital for security engineering, fraud detection, and any system designed to catch deliberate circumvention.

Speaker A: Security is all about detecting evasion.

Speaker B: Exactly. Watch this first brief.

Speaker B: Build a Node.js WAF middleware for an Express API that detects common evasion techniques — including URL encoding tricks, header smuggling, and parameter pollution — logging each evasion attempt with the request fingerprint, IP, and matched rule, and blocking the session after three attempts.

Speaker A: Evasion as a detection category. Very concrete.

Speaker B: Right. Now fraud.

Speaker B: Design a Postgres event schema for a fraud analytics platform that records each evasion pattern identified in payment flows — velocity splitting, micro-transaction probing, device spoofing — with a confidence score, the triggering rule, and the analyst who reviewed it.

Speaker A: Evasion patterns as first-class data objects. Smart.

Speaker B: Now compliance.

Speaker B: Create a Django compliance module for a financial institution that flags accounts showing HMRC-reportable evasion indicators — unexplained offshore transfers, cash structuring below reporting thresholds, and mismatch between declared income and lifestyle transactions — escalating flagged accounts to the AML team.

Speaker A: AML evasion detection built into the compliance workflow.

Speaker B: Now testing.

Speaker B: Write a Playwright test suite for a login system that simulates evasion attacks — CAPTCHA bypass attempts, browser automation fingerprint spoofing, and credential stuffing at rate-limited endpoints — asserting the system blocks each evasion vector within two requests.

Speaker A: Testing your own defences against evasion. Essential.

Speaker B: Last one.

Speaker B: Build a React security dashboard for a SaaS platform that surfaces the top ten evasion techniques attempted against the product in the last thirty days, grouped by technique family, with a timeline of each campaign and a one-click rule-hardening suggestion for each vector.

Speaker A: So evasion in prompts triggers the AI to think adversarially — what will attackers try?

Speaker B: Exactly. Use it any time you want the AI to design for deliberate, intelligent circumvention rather than honest misuse.