Security Theory

Security theory is the part of cybersecurity that tries to state, in precise and checkable terms, what "secure" means.

9 frameworks, from 1949 to today · Cybersecurity · Computer Science

still a working view ended · each leads to its entry below

Overview

Security theory is the part of cybersecurity that tries to state, in precise and checkable terms, what "secure" means. Instead of asking whether a particular system seems safe, it asks what property a system must have, what assumptions that property rests on, and whether those assumptions can be proved or tested. The stakes are practical: a defender who cannot say exactly what they are defending against cannot tell whether a defense works, and a guarantee that rests on an unexamined assumption is no guarantee at all. The history of the subfield is a sequence of attempts to turn vague intuitions—"only authorized users get in," "the attacker cannot learn the message," "the system cannot be subverted"—into formal statements that can be reasoned about, and then to build systems that actually satisfy them.

Secrecy as a Mathematical Problem

The earliest rigorous strand came from cryptography. For centuries ciphers were judged by whether they resisted the attacks people had tried. In 1949 Claude Shannon reframed the question: given a cipher and an adversary with unlimited computing power, how much can the adversary learn1? His answer, perfect secrecy, defined a cipher as perfectly secret when the ciphertext is statistically independent of the plaintext—so seeing the ciphertext changes nothing about the attacker's beliefs. Shannon proved that this requires a key at least as long as the message, which is why the one-time pad is the only perfectly secret cipher. The result was double-edged: it gave secrecy a precise meaning and simultaneously showed that the strongest possible meaning is impractical for most uses. That tension—between provable security and usable security—runs through everything that follows.

The Reference Monitor and the Trusted Computing Base

The next problem was access: how does a system decide which programs and users may read or change which data? In the early 1970s, work associated with the U.S. Air Force and the MITRE Corporation produced the reference monitor concept. A reference monitor is an abstract component that mediates every access by every subject to every object, is always invoked, and is small enough to be analyzed. The concrete implementation of that idea—the hardware and software that must be trusted for the guarantee to hold—became known as the trusted computing base (TCB)2. The insight was structural: security depends on a small, verifiable core, and everything outside it can be untrusted. The 1972 Anderson report, which surveyed computer security for the U.S. government, helped fix these terms in the field's vocabulary. The reference monitor remains the conceptual backbone of operating-system security, and the TCB idea explains why "keep the trusted part small" is a recurring design principle.

Multilevel Security and the Lattice

The reference monitor needed a policy to enforce. The U.S. Department of Defense, concerned with classified information shared among users of different clearances, sponsored a series of models in the 1970s. The best known is the Bell–LaPadula model (1973), which stated two rules: a subject may not read data classified above its clearance (no read up), and may not write data classified below its level (no write down, sometimes called the -property). The second rule prevents a cleared user from leaking secrets into a lower classification. Around the same time, the Biba model (1977) addressed integrity rather than confidentiality, reversing the flow rules to stop low-integrity data from contaminating high-integrity data3.

These models were unified by Dorothy Denning's lattice model of information flow (1976), which treated security labels as elements of a mathematical lattice—a partially ordered set in which any two elements have a least upper bound and greatest lower bound. Information may flow only upward in the lattice. The lattice formulation made multilevel security a general theory rather than a collection of ad hoc rules, and it made the central question explicit: not just who may access what, but where information is allowed to flow.

Formal Verification and the Evaluated System

If a policy can be stated formally, can a system be proved to enforce it? This question drove the most ambitious projects of the 1970s and 1980s. The U.S. Department of Defense's Computer Security Initiative, later the National Computer Security Center, produced the Trusted Computer System Evaluation Criteria (TCSEC), published in 1983 and known as the Orange Book. It defined a ladder of evaluation classes—from D (minimal) through C1, C2, B1, B2, B3, to A1 (formally verified)—each demanding more assurance: more rigorous design, more testing, and at the top, a formal proof that the system's implementation matched its security model. The criteria introduced assurance levels as a graded notion of how much confidence a security claim deserved, distinct from the features a system offered.

The flagship attempt to reach the top was the Honeywell SCOMP (Secure Communications Processor), which in 1985 became the first system evaluated at class A1. Formal verification of real systems proved extraordinarily expensive, and few systems followed. Europe responded with its own schemes—the German Green Book and the British and French criteria—and in the 1990s these were harmonized into the Common Criteria (ISO/IEC 15408, first version 1996, now widely used). The Common Criteria shifted emphasis from a fixed ladder to protection profiles and evaluation assurance levels (EAL1–EAL7), letting a vendor state a target and have it evaluated against that target. The lesson of this era was sobering: formal proof of a whole system is possible but costly, and most practical assurance comes from careful design, testing, and independent evaluation rather than end-to-end proof.

Access Control Beyond the Military

Multilevel security assumed a hierarchical classification. Commercial and civilian systems did not work that way, and the 1980s and 1990s produced models for other settings. Discretionary access control—where an owner decides who may access a file—was the default in early operating systems but offered weak guarantees. Mandatory access control, of which multilevel security is one form, makes access depend on system-wide labels the user cannot change.

A different problem arose in large organizations: permissions multiplied, and administering them individually became unmanageable. Role-based access control (RBAC), formalized by David Ferraiolo and Richard Kuhn in 1992 and later developed into a family of models (RBAC96 by Sandhu and colleagues), assigned permissions to roles rather than to people, and assigned people to roles. This matched how organizations actually worked and made permission management tractable. RBAC became a standard (the NIST model, ANSI INCITS 359) and is now embedded in most enterprise systems. The broader lesson was that access-control theory had to model the organization, not just the data.

Threat Modeling: Asking What the Attacker Will Do

Formal models describe what a system should guarantee; they say little about which threats matter. Threat modeling emerged as the discipline of systematically asking, before building, what an adversary might do. Its modern form was shaped in the late 1990s at Microsoft, where engineers developed a structured process—identify assets, enumerate threats, rank them, and choose mitigations—and published it in the early 2000s. A widely used technique, STRIDE, classifies threats into categories such as spoofing, tampering, repudiation, information disclosure, denial of service, and elevation of privilege4. Threat modeling connected security theory to design practice: it made the adversary's goals an explicit input to the architecture rather than an afterthought.

The Present: Composition, Proof, and the Limits of Models

Today the strands coexist and increasingly interlock. The reference monitor and TCB idea survives in hardware roots of trust and in small verified kernels; seL4, a formally verified microkernel, demonstrated in the 2000s and 2010s that machine-checked proof of enforcement is achievable for a carefully scoped system. Information-flow control has been revived in language-based security, where compilers and type systems track labels through programs. The Common Criteria remains the dominant evaluation framework, though its cost and the difficulty of interpreting assurance levels are persistent criticisms. RBAC has been extended by attribute-based access control, which decides access from arbitrary attributes of user, resource, and environment.

The open questions are the ones the history kept exposing. Perfect secrecy is unattainable in practice, so computational assumptions carry the weight—and those assumptions can fail. Formal proofs are only as good as their models, and models of the adversary are always incomplete. Assurance levels measure process as much as outcome. The field's enduring contribution is not any single model but the habit it established: state the security property precisely, name the assumptions, and make the claim checkable.

References

  1. Communication Theory of Secrecy Systems, Wikipedia
  2. Trusted computing base, Wikipedia
  3. Biba Model, Wikipedia
  4. STRIDE model, Wikipedia

The frameworks of Security Theory

Each framework is a way of working the field has used: when it held, and the question it answers.

  1. 1949-Present
  2. 1972-Present

    Reference Monitor and Trusted Computing Base

    What small core must be trusted to enforce access?

    Fly to its tower in Nousfarer

  3. 1973-1995

    Multilevel security

    How can classified data be shared without leaking secrets?

    Fly to its tower in Nousfarer

  4. 1976-Present
  5. 1983-Present

    Formal Verification and Evaluated Assurance

    Can a system be proved to enforce its security policy?

    Fly to its tower in Nousfarer

  6. 1985-Present

    Access control

    Who may access which resource?

    Fly to its tower in Nousfarer

  7. 1992-Present

    Role-Based Access Control

    How can permissions be managed in large organizations?

    Fly to its tower in Nousfarer

  8. 1996-Present

    Language-Based Information-Flow Control

    Can a program's information flows be checked by its compiler?

    Fly to its tower in Nousfarer

  9. 1999-Present

    Threat model

    What will an attacker actually try to do?

    Fly to its tower in Nousfarer

How the ideas connect

Each arrow runs from an earlier idea to the later one it shaped. The list below says what passed between them.

Perfect Secrecy and Information-Theor…Reference Monitor and Trusted Computi…Multilevel securityLattice-Based Information FlowFormal Verification and Evaluated Ass…Access controlRole-Based Access ControlLanguage-Based Information-Flow Contr…Threat model

Beyond the field

Where Security Theory's ideas came from in other fields, and what they led to there: the strongest 12 of 34 links.

A city in Nousfarer: a red-carpeted stair climbing terraces to a crowned tower on an acropolis, more towers round it, forest and snowy mountains beyond.
A city in Nousfarer, the game played in the browser.

Security Theory is a planet in Nousfarer

Its city has a tower for each of these 9 frameworks. Land, walk in, and climb one concept at a time.

Land on Security Theory

Related fields

Fields its ideas are linked to

More in Cybersecurity

See where Security Theory sits on the Map of Human Knowledge