What malware is and what it wants
Malware is software whose purpose is hostile: steal data, extort money, spy, or recruit your machine into someone else's infrastructure. The classic categories still describe what arrives, even though modern campaigns blend them. Ransomware encrypts files and demands payment, increasingly paired with data theft — pay us or we publish your data — a model called double extortion. Infostealers quietly harvest passwords, session cookies and cryptocurrency wallets, then sell the haul; stolen credentials from these kits feed most follow-on breaches. Spyware watches users, from consumer stalkerware to state-grade commercial spyware. Trojans and backdoors open persistent access for later attackers. Worms spread themselves across networks; botnets enrol devices for spam, click fraud and denial-of-service attacks. Fileless malware lives in memory and trusted scripting tools, leaving few files for scanners to find.
Two details matter more than the taxonomy. First, malware is now an industry: ransomware-as-a-service kits, affiliate programs and initial-access brokers mean the people who break in are often not the people who extort. Second, most successful malware does not "hack" anything exotic — it arrives through a phishing email, a stolen password, or an unpatched known vulnerability. Which is good news, because those are the entries defense can actually close.
How infections actually happen
The delivery routes that dominate real incidents are boring and persistent. Phishing email remains the top door: an attachment (invoice, document with malicious macros) or a link to a credential-harvesting page, increasingly written convincingly because generative AI polished the lures. Unpatched vulnerabilities come next — attackers scan the internet for internet-facing systems running software with known, already-fixed holes; the exploits are often published before organisations patch. Malvertising and poisoned search results place fake download sites above real ones for popular software. Sideloading and pirated software bypass the app-store review that keeps mobile ecosystems comparatively clean. Supply-chain compromise — a vendor's update or dependency gets infected, as in the 2020 SolarWinds case or the 2017 NotPetya outbreak that spread through a Ukrainian accounting suite — is rarer but the most consequential. And removable media still works in industrial settings, a lesson USB-borne incidents keep re-teaching.
Notice what is absent: sophisticated zero-day magic. The highest-volume infections exploit known weaknesses and human trust. That is why the defense list below is unglamorous — and why it works.
Ransomware: the defining threat of the era
Ransomware deserves its own paragraph because it reshaped institutional security. Modern campaigns steal data before encrypting it, so victims face extortion on two fronts; groups operate as businesses with support desks and affiliate programs; and targets shifted from any random PC to hospitals, schools, municipalities and enterprises — organisations whose downtime creates immediate pressure to pay. The landmark incidents are worth knowing: WannaCry (2017) spread globally through an unpatched Windows vulnerability and paralysed parts of the UK's National Health Service; NotPetya (2017), aimed at Ukraine, caused billions in worldwide collateral damage; Colonial Pipeline (2021) halted a major US fuel pipeline and brought ransomware to front pages; and the 2024 international takedown of LockBit showed coordinated law enforcement can disrupt even the largest brands (temporarily — affiliates regroup). The consistent lessons from agency guidance: offline-tested backups, patched internet-facing systems, and never treating the first intrusion as contained until investigated.
Defense that actually works
Patch relentlessly. Enable automatic updates on operating systems, browsers, routers and anything internet-facing; unpatched known vulnerabilities are the single most exploited class. Back up on the 3-2-1 pattern — three copies, two media, one offline or immutable — and test restoration, because untested backups routinely fail exactly when needed. Use phishing-resistant sign-in: hardware keys or passkeys beat SMS codes and defeat most credential theft outright; our online-accounts guide turns this into a checklist. Run reputable endpoint protection with real-time detection — modern tools use behavioural ML to catch unfamiliar malware, which matters as AI accelerates variant production. Limit privileges: everyday accounts should not run as administrators, so malware lands with less room to move. Treat email as hostile terrain: verify senders, distrust urgency, never enable macros because a document told you to, and confirm payment-detail changes out-of-band. Segment and monitor if you run a network: separate backups from the domain, keep logs, and know who to call before the incident. For the AI-specific angle — poisoned assistants and injected instructions — see our operational-AI-security explainer on indirect prompt injection.
If infection is suspected right now: disconnect from the network, run a full scan from a reputable tool, change passwords from a different, clean device starting with email and financial accounts, and if ransomware struck, report to national authorities rather than paying. If personal data was exposed, our breach-response walkthrough covers the first 48 hours step by step.
How AI is changing the malware landscape
Generative AI lowered the cost of convincing lures — flawless localisation, believable tone, rapid variants — and coding assistants can accelerate malware development for actors who already possess criminal intent. On the defense side, machine learning has driven detection for years: behavioural analysis catches never-before-seen samples that signature lists miss, and large-scale models correlate weak signals across millions of endpoints. The equilibrium is shifting rather than collapsing: attacks get cheaper and more personalised; detection gets faster and more adaptive. The durable insight from our AI-in-cybersecurity analysis is that AI favours whichever side has better operational hygiene — patched systems, tested backups, phishing-resistant sign-in — because fundamentals decide outcomes while both sides race. For ongoing coverage of AI-enabled threats and defenses, follow the cybersecurity desk and our AI security & privacy news section.
Frequently asked questions
- What is malware?
- Malware is any software deliberately written to harm a system, its data or its users — steal information, demand payment, spy, recruit the machine into a botnet, or simply destroy. “Virus” is the word most people use, but it names only one type: malware is the umbrella term for viruses, worms, trojans, ransomware, spyware and more.
- What is the difference between a virus and a trojan?
- A virus infects other files and spreads when they are run or shared; a worm spreads by itself across networks without a host file; a trojan disguises itself as legitimate software and relies on the user to install it. Today's major threats — ransomware and infostealers — are usually delivered as trojans or through exploited vulnerabilities rather than classic self-replicating viruses.
- How do I know if my device is infected?
- Common signs: unexplained slowdowns, sudden pop-ups or browser redirects, programs you did not install, files renamed or inaccessible (the classic ransomware symptom), fans running hard while idle, friends receiving messages you did not send, or your antivirus flagging quarantined items. None is proof on its own; a full scan with a reputable, updated scanner is the real test. If ransomware has encrypted files, disconnect from the network immediately and follow an incident guide — our data-breach response walkthrough covers the first steps.
- Should I pay ransomware demands?
- Law-enforcement agencies in the US, UK and EU consistently discourage payment: it funds the criminal ecosystem, does not guarantee working decryption or deletion of stolen data, and marks you as a paying victim. Report the incident to national authorities instead — CISA in the US, the NCSC or Action Fraud in the UK, and national equivalents elsewhere. Recovery without paying is frequently possible from offline backups.
- Are Macs and phones immune to malware?
- No. Platform design differences (app-store review, sandboxing, smaller desktop market share) have historically kept infection rates lower on macOS and mobile, but real malware exists for every major platform — from adware in sideloaded Android apps to cross-platform infostealers. Modern defense is a set of habits, not a specific operating system.
- Is AI making malware worse?
- It is changing both sides. Attackers use generative AI for more convincing phishing lures and to speed up code development; defenders use machine learning to detect novel behaviour faster than signature lists can. The underlying advice does not change — patching, backups, multi-factor authentication and least privilege — but detection cycles on both sides keep shortening, which our AI-in-cybersecurity coverage tracks in detail.