← all systems · see it in The System
Computer security
Computer security (cybersecurity, IT security) is a subdiscipline of information security that protects computer software, systems and networks from threats leading to unauthorized disclosure, theft or damage, or to the disruption or misdirection of their services (src:wp:computer-security). In the field's founding paper, 'security' names the techniques that control who may use or modify the computer or the information in it (Saltzer and Schroeder 1975). Walked here as the field and what it protects: the protection mechanisms, the defenders and attackers, and the contest between them.
Properties published scienceour proposal
Sourced fields, and our own readings.
| born at | 1967 (the Ware session at the Spring Joint Computer Conference); the founding paper on protection 1975 |
| born precision | year; the field had no single birth |
| boundary | Inside: the protection mechanisms (authentication, access control, cryptography, filtering, audit and detection, patching, backups), the people and organisations that defend (security engineers, security operations centres, incident responders, vulnerability researchers, standards bodies), the attackers (threat actors, their malware and infrastructure), and the contest between them. Outside, protected by it: the computers, software, data and networks themselves, the people who use them, and the economy and law that set its incentives. Where the edge with information security (non-computer records) and with physical security sits is not sharp: 'IT security is not a perfect subset of information security' (src:wp:computer-security). |
| nested level | A practice and body of knowledge carried by people and organisations (rung 11-12), running on computers, software and the Internet, which it protects; nested in information security as a subdiscipline; its incentives set by markets and law (Anderson and Moore 2006). |
| scale | technological |
Measures (25) published science
| characteristic clock: Slammer spread (more than 90% of vulnerable hosts infected) | 10 minutes |
| Code Red population doubling time | 37 minutes |
| Slammer: time to full scanning rate | about 3 minutes (more than 55 million scans per second) |
| characteristic clock: human-mediated filtering response to Slammer | within an hour hours |
| Slammer hosts infected | at least 75,000 hosts |
| Slammer per-worm scan rate, Internet-wide average early on | about 4,000 (largest observed 26,000) scans per second per worm |
| time to compromise of an unpatched PC (cited study) | 12 minutes |
| global information security spending, 2026 estimates | 244 to 308 billion US$ |
| malicious share of websites: all vs one vendor's certified | about 3% vs some 8% |
| chance the defender has already found the attacker's bug | 10 % (model: 1000 attacker hours vs 10,000,000 defender hours a year) |
| user-time cost of anti-phishing advice vs phishing losses | two orders of magnitude greater ratio |
| spam: externalities vs spammer's gain | more than 10x ratio (per-hand-off efficiency below 0.1 of social cost arriving as gain) |
| ad fraud: advertiser spend per fraudster dollar | 2-5 US$ per US$ (efficiency 0.2-0.5) |
| share of property crime online | about half by volume and by value |
| ransomware attacks on US health care delivery organisations, 2016-2021 | 374 (PHI of nearly 42 million patients) attacks |
| annual ransomware attacks on US health care, 2016 to 2021 | 43 to 91 attacks per year |
| share of health-care ransomware attacks that disrupted care | 44.4 % |
| computers on the Internet at the Morris worm, 1988 | 60,000 computers |
| password accounts and daily password entries per user | about 25 accounts; 8 per day per user |
| participants fooled by good phishing websites | 90 % |
| DRAM modules with disturbance errors | 110 of 129 modules |
| fewest row accesses to induce a DRAM disturbance error | 139K accesses |
| risk-based alert prioritization vs severity ordering | AUROC 0.92 vs 0.72 (means across eight datasets) |
| Forrest et al.'s computer 'peptide' length | 6 system calls per sequence |
| UK residents kept from online banking by security concerns (Eurostat, as used) | 16 % |
Open contentions (2) open contention
Where sources or readings disagree. Both stay on the map.
- C-1118 In cybersecurity, do attackers have the advantage?
booked: The map keeps both: attack is easier because defenders must cover everything, and attackers have no inherent advantage because one block can break their chain.
the other reading: attacker advantage; no inherent advantage / defender advantage under intelligence-driven defence - C-1125 Does publicly disclosing software flaws make systems safer?
booked: The map follows the finding that disclosure and patching reduce reported flaws, and keeps the argument that disclosure mainly helps attackers alongside.
the other reading: Rescorla: removing an individual bug makes little difference when latent vulnerabilities are many; disclosure aids attackers
Partners and parts published science
Built from / builds (17)
- Cryptography
- Malware
- Authentication
- Access control
- Intrusion detection system
- Security operations center (SOC)
- Incident response
- Vulnerabilities (and their catalogue, CVE)
- Threat actor
- Security policy and governance
- Software vulnerability
- Information security
- Lures against defences for the inbox
- Phishing sites against detection for the golden hours
- Security research against broad access statutes
- Subverters and verifiers over compiler binaries
- RowHammer attacks against memory isolation
Sits inside / holds (23)
- Cryptography
- Cryptographic hash function
- Public-key cryptography
- Human society (from local community up to societies)
- Information security
- Malware
- Authentication
- Access control
- Intrusion detection system
- Security operations center (SOC)
- Incident response
- Threat actor
- Security policy and governance
- Denial-of-service attack
- Firewall (computing)
- Software vulnerability
- Secure software development
- Security information and event management (SIEM)
- Cyber threat intelligence
- Vulnerability management
- Formal security policy model
- Penetration testing
- Security awareness
Trades with (valence) (38)
- Computer memory via sys:electromagnetic-field · Computer security mixed
- Internet protocol suite (TCP/IP): the layered datagram protocols every host and router runs, the RFC record that specifies them, and the community that keeps them via sys:electromagnetic-field · Computer security gain
- Recursion: defining or computing a thing in terms of smaller versions of itself, as computer science and mathematics describe it, with its base cases, call stack, recursive data, induction, and its sightings in language and nature via sys:computer-memory · Computer security mixed
- Central processing unit via sys:electromagnetic-field · Computer security mixed
- Cryptography via sys:electromagnetic-field · Computer security gain
- Cryptographic hash function via sys:written-record · Computer security gain
- Public-key cryptography via sys:electromagnetic-field · Computer security gain
- Human (Homo sapiens) via sys:sensory-environment · Computer security mixed
- Government via sys:written-record · Computer security mixed
- Law (rules and law) via sys:law · Computer security mixed
- Software via sys:written-record · Computer security mixed
- Internet via sys:electromagnetic-field · Computer security mixed
- Immune system via sys:written-record · Computer security gain
- Money via sys:money · Computer security loss
- Computer via sys:electromagnetic-field · Computer security gain
- Market and trade via sys:money · Computer security mixed
- Health care via sys:written-record · Computer security mixed
- Artificial intelligence via sys:written-record · Computer security mixed
- Compiler via sys:software · Computer security mixed
- Authentication via sys:electromagnetic-field · Computer security gain
- Access control via sys:electromagnetic-field · Computer security gain
- Intrusion detection system via sys:electromagnetic-field · Computer security gain
- Security operations center (SOC) via sys:human-society · Computer security gain
- Incident response via sys:electromagnetic-field · Computer security gain
- Threat actor via sys:computer-network · Computer security mixed
- Security policy and governance via sys:organization · Computer security gain
- Phishing via sys:internet · Computer security gain
- Information security via sys:written-record · Computer security gain
- Computer hardware via sys:electromagnetic-field · Computer security mixed
- Database via sys:written-record · Computer security gain
- Computer network via sys:electromagnetic-field · Computer security gain
- Software vulnerability via sys:written-record · Computer security mixed
- Secure software development via sys:software · Computer security gain
- Social engineering via sys:electromagnetic-field · Computer security loss
- Formal security policy model via sys:computer-security · Computer security gain
- Security awareness via sys:sensory-environment · Computer security gain
- Computer ethics via sys:computer-security · Computer security gain
- Computer law via sys:electromagnetic-field · Computer security mixed
expressed by (31)
- Recombination/Composition
- Templating / Standardization
- Encapsulation & Modularity (Abstraction)
- Internal Coordination/Regulation
- Sensing/Scanning
- Mapping/Modeling
- Uncertainty Reduction/Learning
- Active Experimentation / Probing (Optimal Experimental Design)
- Trail-Marking / Externalized Memory (Stigmergy)
- Surprise / Change-Point Detection
- Social Learning / Public Information Use
- Evade Detection
- Overcome Barrier
- Speed
- Resource Denial/Attrition
- Subversion & Co-optation
- Prepositioning & Latency
- Leverage Host/Amplification
- Cascade Induction
- Reactive Response
- Proactive Defense
- Physical/Virtual Barrier
- Technique-Based Identification (a.k.a. Technique-Based Detection)
- Self-Monitoring
- Validation of Identity/Intention
- Enclave/Segregation
- Deception
- Resilience & Repair
- Redundancy & Diversity
- Triage & Prioritized Allocation
- Quality Control & Error Correction
floor filled by (19)
- Cryptography
- Public-key cryptography
- Human (Homo sapiens)
- Law (rules and law)
- Software
- Internet
- Human society (from local community up to societies)
- Computer
- Market and trade
- Authentication
- Access control
- Intrusion detection system
- Security operations center (SOC)
- Incident response
- Security policy and governance
- Information security
- Software vulnerability
- Secure software development
- Penetration testing
same practice as (1)
Kinds (5)
- Cryptography
- Information security
- Network security
- Application and software security (secure coding)
- Hardware security
wired (1)
Notes published science
- A worm that didn't wait won: Code Red opened a connection and waited for each reply, doubling every 37 minutes; Slammer fit its whole self into one 404-byte packet that needed no reply, so it was limited only by bandwidth, two orders of magnitude faster, and infected over 90% of vulnerable servers in 10 minutes, long before any person could react (src:moore-2003-slammer).
- Protection is often worth less than it costs the people asked to do it: a minute a day spent checking web addresses would cost users about a hundred times more than all phishing losses, so ignoring the advice is rational (src:herley-2009-externalities); and a spam campaign burns more than ten dollars of other people's value for each dollar the spammer makes.
- Every computation leaks into the world around it, and the leak becomes someone else's sense: precise timing of key operations can give away a secret key (src:kocher-1996-timing), and reading one row of memory over and over leaks charge into the neighbouring rows until their bits flip, in 110 of 129 memory modules tested (src:kim-2014-rowhammer).
- Bell's 'Computer Security Intermediate Value Theorem': connect a computer at one classification level to one at another through any network cloud, and some box inside the cloud must be handling both levels. You can move the hard part, 'but you cannot eliminate it' (src:bell-2005).
Layers
The ways it takes in its world, where they meet, and how it acts back, each from L1 (where it meets the world) up.
the attacker's side: from the target's exposed surface up to the motive (secondary stack)
- L1: The target's exposed surface: addresses, listening services, people, physical leaks
- L2: Probes, scans and measurements
- L3: Replies and leaks read; the defender's detectors modelled
- L4: The operation: weaponization, delivery, exploitation, installation, command and control, actions on objectives
- L5: Campaigns: intrusions linked by reused tools; persistence
- L6: Criminal markets and infrastructure (botnets, crime services)
- L7: Motives and backers
the boardroom (shared floors)
- L4: Detection and analysis: security operations centres screen alerts (filter, prioritize, correlate, contextualize) under large numbers of false alerts; NIST DETECT
- L4: The attacker's operation: the intrusion kill chain from reconnaissance through weaponization, delivery, exploitation, installation and command and control
- L5: Response and recovery: incident response contains, eradicates and restores; NIST RESPOND and RECOVER
- L5: The attacker's campaign: intrusions linked by reused tools and infrastructure; persistence over months
- L6: Risk understanding, policy and models: risks understood (NIST IDENTIFY), security policy and threat models, the economics of incentives, work factor
- L6: Criminal markets and infrastructure: botnets, pay-per-install, crime-as-a-service; grey markets for exploits
- L7: Governance and purpose: strategy, expectations and policy set by the people the field serves; the security goals (confidentiality, integrity, availability)
- L7: Motives and backers: thrill, activism, profit, state sponsorship
the protection of computing: the world it touches up to the goals people set for it (reader's view)
- L1: The physical world computing runs in: hardware, signals, people at keyboards; every computation affects its surroundings
- L2: The protected assets: computers, software, data and networks under controlled sharing, with their vulnerabilities
- L3: Protection mechanisms: authentication, access control with complete mediation, cryptography, filters; defence in depth (NIST PROTECT)
- L4: Detection and analysis: audit records, intrusion detection, alert screening in security operations centres (NIST DETECT)
- L5: Response and recovery: incident response contains and restores (NIST RESPOND, RECOVER)
- L6: Risk, policy and models: risks understood (NIST IDENTIFY), threat models, work factor, the economics of incentives
- L7: Governance and goals: confidentiality, integrity and availability, and the strategy and policy that set them (NIST GOVERN)
way in: Credentials presented at the gate
- L1: A person or agent making a request, holding secrets in memory
- L2: Credentials typed (about 25 accounts, 8 password entries a day per user)
- L3: Verification of identity; every access checked for authority
way in: Host and application events (audit records, system-call traces)
- L1: The protected computers' activity: logins, file accesses, system calls of privileged processes
- L2: Audit records and logs generated and kept (stored remotely, append-only)
- L3: Profiles of normal behaviour and anomaly rules; signatures of known attacks
way in: Losses and costs (the economic channel)
- L1: Direct losses, indirect losses and defence costs
- L2: Incidents counted and reported (e.g. ransomware on health care: 374 attacks, nearly 42 million patients' records)
- L3: Aggregated estimates and victimisation studies
- L4: Priced by insurers (correlated risks hard to insure)
- L5: Market reaction: share prices fall after breaches that leak confidential information
way in: Malware samples
- L1: Malicious programs released into cyberspace
- L2: Samples collected (honeypots, user submissions, sensors)
- L3: Classification by signatures and machine-learning detectors with chosen feature sets
way in: Network traffic (packets at sensors and filters)
- L1: Packets crossing an organisation's links, legitimate and hostile
- L2: Routers, firewalls and network sensors that see and can filter traffic
- L3: Signatures that distinguish attack traffic
way in: People as sensors (users noticing and reporting)
- L1: A message or page in front of a person (Human channel sight)
- L2: Noticing cues: address bar, status bar, security indicators
- L3: The vigilant user as a detector at the delivery phase
way in: Reconnaissance (the attacker's view of the target)
- L1: The exposed surface: addresses, listening services, people
- L2: Probes and scans
- L3: Replies read, or no reply needed (the wait)
way in: Side channels (what a computation leaks)
- L1: A computation's physical effects on its surroundings
- L2: Measurement of timing (or power, emanations)
- L3: Inference of internal state (keys) from the measurements
way in: The defender as seen by the attacker
- L1: Deployed detectors, filters and their decisions
- L2: Knowledge of the defender's feature set
- L3: Evasive manipulation planned
way in: Threat intelligence (indicators shared between defenders)
- L1: Intrusions seen elsewhere (other organisations, earlier campaigns)
- L2: Indicators extracted: atomic, computed, behavioral
- L3: Indicators matured into tools and matched against activity
way in: Vulnerability discovery and disclosure
- L1: Flaws in software and hardware (about 1% of a large program's bugs security-critical, in Anderson's model)
- L2: Discovery by testing, research and attackers (a race; vulnerabilities often rediscovered)
- L3: Disclosure and cataloguing (CVE, advisories)
- L4: Pricing and market for vulnerability reports
- L5: Vendor response: patches (faster under public disclosure)
way out: Actions on objectives (theft, ransom, disruption, physical damage)
- L1: Effects in the world: downtime, cancelled care, ambulance diversion; centrifuges
- L4: Actions on objectives, the last kill-chain phase
- L7: The motive picks the objective
way out: Advice and training to users
- L1: Advice reaches people, who often rationally reject it
- L6: Security advice issued
way out: Attack delivery (exploit, install, command and control)
- L1: A packet or file crosses into the target (Slammer: one UDP packet)
- L4: Delivery and exploitation chosen
way out: Containment and filtering
- L1: Hostile packets dropped (and legitimate ones, if the port is shared)
- L3: Filter rules on routers and firewalls
- L5: Responders decide to contain
way out: Deception (honeypots, tarpits)
- L1: Decoy computers left vulnerable on purpose
- L5: Responders choose deception courses of action
way out: Recovery (restore from backups)
- L1: Data and systems restored (in health care, less often from backups over 2016-2021)
- L5: Recovery restores assets and operations
way out: Safeguards deployed (access control, encryption, patches)
- L1: Patches and configuration written into the computers
- L3: Mechanisms: every access checked; least privilege; encryption
- L6: Safeguards chosen against risks
- L7: Policy sets what is protected and how much
way out: Sharing and disclosure (indicators, advisories, reports)
- L1: Advisories and patches published; read by both sides
- L6: Indicators revealed through analysis or collaboration
way out: Social engineering and phishing (attacks on people)
- L1: Imitation pages and messages on a person's screen (Human sight; urgency on emotion)
- L4: Pretexts chosen: impersonation, urgency
Who fills each layer
- L3: Cryptography
- L4: Intrusion detection system, Cyber threat intelligence
- L5: Incident response
- L6: Penetration testing, Formal security policy model