{
  "id": "ransomware",
  "title": "Ransomware",
  "lang": "en",
  "source_lang": "en",
  "summary": "Ransomware is malicious software used to extort payment by denying access to systems or data, often combined with threats to disclose stolen information.",
  "category": "computer-science",
  "subcategory": "networks-security",
  "type": "concept",
  "aliases": [
    "Ransom malware",
    "Crypto-ransomware"
  ],
  "generated_at": "2026-11-02",
  "model": "pro/gpt-6.1-sol",
  "url": "https://aiwiki.page/en/ransomware/",
  "translations": {
    "zh-Hans": {
      "url": "https://aiwiki.page/zh/ransomware/",
      "json": "https://aiwiki.page/zh/ransomware.json"
    }
  },
  "sections": [
    {
      "heading": "",
      "text": "Ransomware is a class of malicious software that makes files, applications, or computer systems unavailable and demands payment for restoring access. Modern ransomware commonly uses encryption to render data unreadable without a decryption key. Many campaigns also steal information and threaten to publish it, combining disruption with coercion over confidentiality. In broader usage, “ransomware attack” can describe the entire extortion operation, including intrusion, data theft, negotiations, and recovery, rather than only the program that encrypts files.\n\nRansomware is a major concern in cybersecurity because a compromise of ordinary business systems can interrupt hospitals, municipal services, manufacturing, transportation, and other activities dependent on digital infrastructure. Its development reflects the interaction of cryptographic techniques, network vulnerabilities, criminal specialization, and digital payment systems. Understanding it therefore requires distinguishing the software’s technical mechanisms from the wider economic and organizational conditions that make extortion possible."
    },
    {
      "heading": "Definition, boundaries, and forms",
      "text": "The defining feature of ransomware is coercive denial of access coupled with a demand for payment. In an encryption-based attack, files remain present but their contents become unusable. In a screen-locking attack, the software obstructs interaction with the device, sometimes without altering the underlying files. Other variants interfere with startup, storage access, or applications. The scope ranges from a single personal computer to thousands of servers and workstations managed by one organization.\n\nRansomware attacks affect different objectives of information security. Encryption primarily compromises availability, while unauthorized modification compromises integrity. Data theft compromises confidentiality even if no encryption occurs. This distinction matters because the remedies differ: restoring files from a clean backup can address loss of availability, but it cannot retract information already copied by an attacker. Similarly, possessing a working decryption key does not establish that the attacker has left the network.\n\nThe term crypto-ransomware refers specifically to ransomware that encrypts data. It does not mean that the malware necessarily demands cryptocurrency. Locker ransomware generally blocks use of a device or session, while scareware relies on false warnings or fabricated problems to induce payment. These categories overlap in practice, but software that merely displays a fraudulent warning is technically different from software that has cryptographically transformed the victim’s documents. Some attacks also target a database or network-attached storage device rather than desktop files.\n\nModern extortion terminology describes additional pressure mechanisms. Double extortion usually means combining encryption with a threat to disclose stolen data. Triple extortion has no single standardized meaning; it may refer to adding denial-of-service attacks, pressure on customers or employees, or another coercive measure. An operation based entirely on stolen-data threats is more precisely described as data extortion, although industry reporting often groups it with ransomware. Such differences in terminology complicate comparisons between incident counts.\n\nA further boundary separates ransomware from destructive malware that only pretends to offer recovery. A wiper is designed to destroy data or make it inaccessible without a genuine restoration mechanism. A ransom note alone does not prove that an incident is financially motivated or that payment could recover the affected data. Conversely, an unsuccessful ransomware program may destroy files accidentally through programming errors. Classification therefore depends on technical behavior and operational evidence, not simply the message shown to victims."
    },
    {
      "heading": "Historical development and major incidents",
      "text": "The AIDS Trojan of 1989 is widely recognized as an early ransomware example. Distributed on floppy disks associated with AIDS information, it altered file names and hid directories after a period of use, then demanded payment through instructions referring to a post-office box in Panama. Its concealment method was comparatively weak and recoverable through technical analysis. Nevertheless, it established the central pattern of using software to withhold access and demand money for restoration.\n\nA more powerful theoretical foundation appeared in 1996, when Adam Young and Moti Yung described cryptovirology: the offensive use of cryptography to give an attacker capabilities that defenders could not readily reverse. Public-key techniques made it possible for malicious software to encrypt data without containing the corresponding secret needed to decrypt it. The growth of the Internet, electronic payment channels, and large populations of networked computers later supplied an environment in which that concept could support scalable criminal operations.\n\nDuring the 2000s, ransomware included fake antivirus products, screen lockers, and programs impersonating police agencies. Some claimed that a computer had been locked because of illegal activity and demanded a supposed fine. CryptoLocker, first observed in 2013, became an influential example of large-scale file-encrypting ransomware supported by public-key cryptography. Its distribution was associated with the Gameover Zeus criminal infrastructure. A coordinated international operation disrupted that infrastructure in 2014, illustrating how action against supporting services could affect a ransomware campaign.\n\nTwo outbreaks in 2017 demonstrated different forms of systemic risk. WannaCry spread in May by exploiting a vulnerability in the Windows implementation of the Server Message Block protocol, using the EternalBlue exploit. Microsoft had issued the relevant security update, MS17-010, in March. Its worm-like propagation caused disruption across many countries, including substantial effects on England’s National Health Service. NotPetya followed in June, initially spreading through a compromised update mechanism for the Ukrainian accounting software M.E.Doc. Although it displayed a ransom demand, its recovery design and destructive behavior led it to be widely classified as a destructive attack rather than a conventional revenue-seeking ransomware campaign.\n\nFrom the late 2010s, many operators shifted toward manually directed attacks on organizations rather than indiscriminate encryption of individual computers. Maze helped popularize combined encryption and public data-leak threats in 2019. In May 2021, the Colonial Pipeline incident showed how compromised business systems could precipitate precautionary shutdowns affecting fuel distribution even without publicly established encryption of pipeline control equipment. The July 2021 Kaseya VSA incident demonstrated another scaling mechanism: compromising remote-management software used by service providers to reach multiple downstream organizations."
    },
    {
      "heading": "Encryption, keys, and software behavior",
      "text": "Effective crypto-ransomware generally uses a hybrid cryptographic design. A fast symmetric algorithm encrypts file contents, while an asymmetric mechanism protects the symmetric keys. Encrypting large amounts of data directly with public-key operations would be inefficient; using symmetric encryption for the bulk data combines speed with a mechanism that can keep the recovery secret outside the victim’s reach. Implementations vary substantially, and some use key-agreement techniques rather than directly encrypting a file key with a public key.\n\nIn a simplified scheme, a file \\(M\\) is encrypted with a randomly generated key \\(K\\), producing ciphertext \\(C\\). The attacker’s public key \\(pk_A\\) then protects \\(K\\), producing a wrapped key \\(W\\):\n\n\\[C=\\operatorname{Enc}_{K}(M), \\qquad W=\\operatorname{Enc}_{pk_A}(K).\\]\n\nThe corresponding private key is retained by the attacker. A real implementation may use separate keys for files, directories, devices, or entire campaigns. Depending on the cryptographic mode, it also needs correctly managed initialization vectors or nonces. When the design is sound and secret key material is unavailable, reverse-engineering the executable does not automatically reveal a practical way to decrypt the files.\n\nRansomware often favors speed over encrypting every byte. Some families encrypt selected portions of large files, alternate encrypted and unencrypted regions, or concentrate on file headers and other structures essential to interpretation. This is sometimes called intermittent encryption. A relatively small amount of corruption may make a large document, virtual disk, or database unusable. However, the relationship between encrypted volume and recoverability depends on file format and application behavior; encrypting a fixed fraction of bytes does not guarantee a fixed amount of damage.\n\nThe surrounding software behavior can be as consequential as the cipher. An encryptor may terminate applications so that their files are no longer locked, interfere with local recovery facilities, enumerate network shares, change extensions, and place ransom notes in affected directories. Some programs avoid selected system files so that the operating system remains functional enough to display payment instructions. Others target virtualization hosts, where encrypting a small number of storage files can disable many virtual machines at once. Versions exist for Microsoft Windows, Linux, and other environments, but functionality differs among families.\n\nCryptographic mistakes can create recovery opportunities. Historical cases have involved predictable keys, weak randomness, reused secret material, incomplete encryption, or keys left in memory or other recoverable locations. Investigators have also obtained keys through infrastructure seizures or disclosures. None of these possibilities makes universal decryption feasible. A decryptor for one family or version may be ineffective against another, and even a legitimate decryptor may not repair corrupted files, recover deleted data, or restore application consistency."
    },
    {
      "heading": "Intrusion, lateral movement, and deployment",
      "text": "In many organizational incidents, ransomware is the final stage of a broader intrusion rather than the initial means of entry. Attackers may enter through phishing, compromised credentials, exposed remote-access services, or vulnerabilities in Internet-facing applications. Malicious email attachments remain one route, but not all ransomware begins with a user opening a file. A vulnerable edge appliance or an account obtained through a separate credential-stealing campaign can provide access without any interaction by the eventual victim organization’s staff.\n\nFollowing entry, attackers often seek greater privileges and a clearer view of the environment. They may identify directory services, file servers, backup systems, security products, and administrative accounts. Lateral movement describes the use of access on one system to reach others. The same remote-management mechanisms that support legitimate administration can also support unauthorized movement when an attacker controls a sufficiently privileged identity. Consequently, an incident may contain little unusual software until relatively late in the intrusion.\n\nIdentity infrastructure has particular importance. Active Directory, widely used to organize Windows domains, centralizes many administrative relationships. Its compromise can permit widespread software deployment or access to large numbers of systems. The crucial issue is often control over authorized management functions rather than exploitation of a new vulnerability on every machine. This explains why a single stolen administrative credential can have consequences disproportionate to the number of devices initially compromised.\n\nData theft may precede encryption by hours, days, or longer. Attackers collect information, package it, and transfer it to infrastructure under their control. The files actually taken may differ from the volumes later claimed in a ransom note or leak-site announcement. Encryption is then scheduled to maximize operational pressure or to affect many systems in a short period. Some intrusions are rapid, while others involve extended reconnaissance; no fixed sequence or dwell time applies to all ransomware operations.\n\nPropagation mechanisms also vary. A computer worm spreads through automated replication, as WannaCry did under suitable network conditions. A human-operated campaign instead relies on an intruder’s decisions and administrative control, although deployment itself may be automated. Supply-chain and managed-service-provider compromises introduce another pattern, in which access to a trusted supplier becomes a route into customers. In cloud computing, comparable extortion can involve deletion of resources, modification of access controls, or interference with backups rather than execution of a conventional file encryptor. These distinctions influence both the evidence left behind and the extent of the affected environment."
    },
    {
      "heading": "Criminal business models, payment, and negotiation",
      "text": "Ransomware operations frequently involve a division of labor. An initial-access broker sells access to compromised organizations; an affiliate conducts intrusions; a developer maintains the encryptor; and other participants operate infrastructure, negotiate payments, or move proceeds. These roles can overlap, and a single operator may perform several of them. The visible ransomware brand is therefore not necessarily a stable organization with a fixed membership.\n\nRansomware as a service, commonly abbreviated RaaS, formalizes some of this specialization. An operator supplies software, payment infrastructure, administrative panels, or leak-site services to affiliates in exchange for a share of proceeds or another payment arrangement. The model resembles software as a service in its division between platform provision and use, but it is a criminal commercial arrangement rather than a standardized technical architecture. Affiliates may move between brands, and a disruption of one brand need not eliminate the people or access networks associated with it.\n\nCryptocurrency became important because it supports cross-border transfers without the conventional card-payment mechanisms that can readily be reversed. Bitcoin has been prominent, although some operators have requested other currencies, including privacy-oriented ones. Cryptocurrency does not automatically provide anonymity. Public blockchain records can expose transaction relationships, while exchanges and other services may hold identifying information. At the same time, intermediary wallets, mixing services, and cross-jurisdictional transfers can complicate attribution and recovery.\n\nNegotiations commonly use dedicated portals or messaging channels. Operators may present samples of stolen files, offer limited demonstration decryption, impose deadlines, or change demands over time. A demonstration establishes only that particular sample files can be decrypted; it does not prove that every affected system is recoverable or that stolen information will be destroyed. Ransom demands may reflect the attacker’s estimate of the organization’s resources, operational dependence on affected systems, insurance position, and likely cost of prolonged disruption.\n\nThe economic decision is not simply a comparison between a ransom and the price of restoring a backup. Disruption costs continue during negotiation and recovery, while payment creates no enforceable guarantee of performance. A simplified analytical model may represent expected direct cost as\n\n\\[E[C]=R+E[L]+E[F],\\]\n\nwhere \\(R\\) is a ransom payment if made, \\(L\\) is interruption and recovery loss, and \\(F\\) represents additional costs such as investigation and legal obligations. The terms are uncertain and interdependent: payment may fail to reduce interruption, and functioning backups may leave disclosure threats unresolved. Aggregate estimates of criminal revenue are similarly incomplete because private settlements, unidentified wallets, and unreported incidents remain outside many datasets."
    },
    {
      "heading": "Operational consequences and measurement difficulties",
      "text": "The immediate effect of ransomware is often a loss of access to information or computing services. An organization may be unable to authenticate users, process orders, schedule work, issue invoices, or retrieve records. Disruption can also arise from containment decisions: systems that were not encrypted may be disconnected because they share a network or identity infrastructure with compromised systems. The operational footprint of an incident can therefore exceed the footprint of the malware itself.\n\nSectoral consequences depend on the relationship between business computing and physical activity. Manufacturing may stop when scheduling, inventory, or quality-control systems become unavailable. Local governments can lose access to permitting, tax, or administrative services. Healthcare organizations have reported postponed appointments, diverted patients, and temporary reliance on paper processes. Such consequences establish a serious operational concern, but attributing a particular clinical outcome to ransomware requires evidence beyond the fact that an attack occurred.\n\nRecovery costs also extend beyond replacing computers or obtaining a decryptor. Organizations may need to reconstruct identities, rebuild servers, verify backups, reconcile transactions, and establish whether restored information is complete. Lost sales, contractual penalties, notification expenses, litigation, and reputational effects may follow. An attack on a provider serving many customers can spread these costs across organizations that were never directly targeted. Conversely, some systems may be restored quickly while specialized applications take much longer because they depend on tightly coupled data and configuration.\n\nReliable measurement is difficult because available datasets observe different populations. Police reports capture reported crimes; insurer data reflect insured organizations; incident-response firms observe their own clients; cryptocurrency analysis follows identifiable transactions; and leak sites record an attacker-selected subset of cases. A leak-site posting is evidence of a claim, not independent verification of the quantity or sensitivity of stolen information. Counting postings also misses victims who are never listed and may count the same organization more than once across reporting systems.\n\nFinancial statistics require similar caution. A ransom demand is not the same as a payment, and ransom payments are not the same as total losses. The US Federal Bureau of Investigation’s Internet Crime Complaint Center reported 2,825 ransomware complaints and more than $59.6 million in adjusted losses for 2023. Its ransomware figures do not represent a complete total of business interruption, lost productivity, recovery expenses, or all unreported incidents. Comparisons across years or countries can also reflect changing reporting practices, classifications, and detection coverage rather than a corresponding change in the underlying incidence of attacks."
    },
    {
      "heading": "Prevention, detection, and resilience",
      "text": "Ransomware defense is usually described as a layered problem because no single control addresses every stage of an attack. Vulnerability management reduces exposure to known software flaws; secure configuration limits unnecessary access; and strong authentication makes stolen passwords less useful. Multi-factor authentication can reduce certain credential-based compromises, but its effectiveness depends on the implementation and the route being attacked. Session theft, unauthorized account recovery, and compromised administrative endpoints illustrate why identity security involves more than adding a second login factor.\n\nPrivilege separation and network segmentation seek to limit the consequences of initial compromise. A workstation account need not possess authority over backup infrastructure, and a system serving one business function need not be able to administer every other system. Such boundaries can restrict lateral movement and mass deployment. Their actual effectiveness depends on the credentials, trust relationships, and management channels that cross them. A nominally separate network may still share an administrative dependency that enables an attacker to reach both environments.\n\nDetection combines endpoint, network, identity, and storage evidence. Potential signals include unusual privilege changes, unexpected administrative connections, bulk modification of files, attempts to disable security services, and interference with recovery mechanisms. Anomaly detection can help identify departures from ordinary activity, but legitimate migrations, backup jobs, and software deployments can produce similar patterns. Detection quality therefore involves both sensitivity and context, rather than simply classifying every high-volume file operation as malicious. Waiting for encryption to begin may also miss an earlier opportunity to detect data theft.\n\nBackups are central to resilience but are not equivalent to guaranteed recovery. A backup reachable with the same compromised credentials as production data may be altered or deleted during the intrusion. Offline, isolated, or appropriately immutable copies introduce different barriers, although their protection depends on configuration and administrative control. Recovery also requires usable copies of application data, configurations, encryption keys, and other dependencies. A successful file-level restore is not proof that an entire business service can resume.\n\nOrganizations commonly distinguish the recovery point objective, which concerns tolerable data loss measured in time, from the recovery time objective, which concerns the duration of service interruption. Tested restoration procedures establish whether these objectives are realistic under actual storage, staffing, and application constraints. Resilience also includes alternative operating procedures and prioritization of essential services. These measures address disruption, while separate controls govern sensitive-data access and exfiltration; restoring from backup does not resolve a confidentiality breach or invalidate copies held by an attacker."
    },
    {
      "heading": "Incident response, law enforcement, and unresolved challenges",
      "text": "Incident response treats ransomware as an active compromise whose boundaries must be established. The visible ransom note may appear only after credentials have been stolen, information copied, and additional access mechanisms installed. Investigators therefore distinguish immediate containment from full eradication. Disconnecting a compromised system can reduce further spread, but it does not by itself identify every affected account, connected service, or external transfer of data.\n\nDigital forensics reconstructs the intrusion from sources such as endpoint artifacts, authentication records, network logs, cloud audit trails, and the malware itself. Evidence may establish the entry point, deployment method, affected files, and indicators of exfiltration. Its absence must be interpreted cautiously: an attacker may delete logs, retention periods may be short, and some activities leave limited traces. A finding that no exfiltration was observed is not automatically proof that none occurred. Rebuilding a trustworthy environment can consequently require credential replacement and administrative reconstruction in addition to decrypting or restoring files.\n\nLaw-enforcement operations target multiple parts of the ecosystem, including infrastructure, criminal marketplaces, payment flows, and individuals. The 2021 operation against Emotet infrastructure and the 2023 disruption of Qakbot addressed malware services that had supported wider criminal activity, including ransomware delivery. In February 2024, Operation Cronos disrupted LockBit infrastructure through an international law-enforcement effort. Such actions can produce seizures, intelligence, and decryption assistance, but the long-term effect depends partly on whether participants can reconstitute services or migrate to other criminal networks.\n\nLegal consequences differ by jurisdiction and by the facts of an incident. Data-protection laws may require notification of qualifying personal-data breaches; sectoral rules may impose additional reporting duties. In the European Union, the General Data Protection Regulation includes notification requirements tied to the nature and risk of a personal-data breach. Payment can also raise sanctions issues where a recipient is designated under applicable law. The United States Treasury’s Office of Foreign Assets Control has published advisories concerning the potential sanctions risks of facilitating ransomware payments. Neither payment nor successful restoration necessarily removes reporting or regulatory obligations.\n\nSeveral technical and policy questions remain unresolved. Researchers study how to detect malicious encryption without excessive false alarms, how to maintain recoverable systems when administrative identities are compromised, and how to measure attacks that are frequently concealed or incompletely reported. Attribution presents a separate difficulty: shared code, reused tools, affiliate movement, and deliberately misleading artifacts complicate identification of the people behind a campaign. Evaluating payment restrictions, mandatory reporting, insurance conditions, and international disruption efforts requires distinguishing their effects on observed reporting from their effects on actual attacks, criminal revenue, and victim harm."
    }
  ],
  "links": [
    "malware",
    "encryption",
    "cybersecurity",
    "personal-computer",
    "information-security",
    "database",
    "cryptography",
    "internet",
    "algorithm",
    "operating-system",
    "virtual-machine",
    "microsoft-windows",
    "linux",
    "phishing",
    "email",
    "cloud-computing",
    "software-as-a-service",
    "authentication",
    "anomaly-detection",
    "incident-response",
    "european-union"
  ],
  "sources": [],
  "license": "https://creativecommons.org/licenses/by/4.0/"
}