Quira Quira Docs
GitHub Star

Security Overview

Quira is designed from the ground up with security as a first-class concern. But Quira goes further than any existing browser: it introduces Context Security — a new security discipline that protects your accumulated knowledge structure, not just your network connection.

Context Security: a new paradigm

Every major browser — Chrome, Firefox, Brave — focuses on protecting users from the web via sandboxing, site isolation, and content security policies. None of them protect the user's own accumulated knowledge with the same rigor.

Quira's Context Graph is an automatically constructed personal knowledge graph — a structured map of your intellectual activity. This is an entirely new category of sensitive data that no existing browser security model protects. Context Security is the discipline of protecting personal knowledge structures.

Why this matters

The Context Graph holds AI summaries, entity extractions, semantic embeddings, browsing patterns, and temporal research paths. This is categorically more sensitive than browsing history — it is a machine-readable map of your thinking.

Compositional sensitivity

Individual page visits are low-sensitivity data. But when accumulated in the Context Graph — combined with AI summaries, extracted entities, and temporal patterns — they become one of the most sensitive personal datasets possible. This is the same principle by which individual mobile location pings are harmless, but aggregated location history becomes complete surveillance.

Eight security paradigms

Quira's security architecture is built on eight interlocking paradigms. Each is independently valuable; together they form a living, adaptive, cryptographically-grounded defense system.

#ParadigmCore ideaDetails
1Capability-Based SecurityUnforgeable tokens replace identity-based access. No ambient authority.Permission System
2Information Flow ControlTrack where data flows, not just who can access it. Labels are monotonic.Permission System
3AI Immune SystemBehavioral baselining and anomaly detection. The AI defends itself.Threat Protection
4Temporal SecurityPermissions decay. Sessions are time-locked. Dead man's switch.Privacy Architecture
5Adversarial ML DefenseMulti-layer defense against web content manipulating browser AI.Threat Protection
6Cryptographic InnovationForward secrecy at rest, threshold encryption, verifiable deletion, ZKP.Privacy Architecture
7Hardware Co-DesignTEE for AI inference, memory encryption, per-tab decryption.Privacy Architecture
8Living SecuritySecurity Health Score, continuous adaptation, not static configuration.Threat Protection

Beyond these eight paradigms, Quira introduces 26 novel security structures — from macOS deep-architecture concepts (8 structures including CSEB, KPC, CTA) to emergent interaction structures (8 structures including KSCI, IRD, RSIP, CAV). See each page for formal models and implementation details.

Defense in depth

Quira implements a defense-in-depth strategy inspired by macOS's multilayered security model. Rather than relying on a single boundary, multiple independent layers ensure that a breach in one does not compromise the whole system.

macOSQuira equivalentDifference from existing browsers
GatekeeperCryptographic extension verification before installChrome reviews after publish; Quira verifies before execution
TCCContext TCC — granular per-permission consent for graph accessWorld first: no browser has knowledge graph access control
SIPContext Graph immutability protectionNo browser guarantees browsing data integrity
App SandboxSpace-Scoped Security — Context Spaces are security boundariesWorld first: tab groups as security boundaries
KeychainEncrypted embedding vault with independent encryptionWorld first: encrypted browsing data at rest
Lockdown ModeContext Lockdown — JIT disabled, panic wipe, RAM-only modeWorld first: extreme privacy mode for journalists/activists

Security layers

LayerComponentWhat it protects
L1 — NetworkNetwork FirewallMalicious connections, DNS filtering, TLS enforcement
L2 — ContentContent FilteringTrackers, ads, fingerprinting scripts, cryptominers
L3 — ThreatsThreat ProtectionPhishing, malware, AI immune system, adversarial ML defense
L4 — PermissionsPermission SystemCapability tokens, Context TCC, IFC labels, Space isolation
L5 — PrivacyPrivacy ArchitectureEncryption at rest, temporal security, hardware integration
L6 — InfrastructureAdvanced ArchitectureKPC classification, CSEB event bus, KQR relay, CTA trust arbiter, ASP posture
L7 — EmergentEmergent SecurityKSCI supply chain, IRD inference residue, SBRC blast radius, TCV causality, CAV amnesia

Competitive comparison

FeatureChromeFirefoxBraveTorQuira
Site isolationYesYesYesYesYes (Gecko Fission)
Extension sandboxBasicBasicBasicRestrictedCapability-based TCC
History encryptionNoNoNoRAM onlySQLCipher + embedding encryption
AI threat defenseNoNoNoN/A4-layer prompt injection defense
Knowledge exfiltration detectionNoNoNoNoContext DLP
Lockdown modeNoNoNoPartialContext Lockdown
Compositional sensitivityNoNoNoNoYes — world first

Implementation roadmap

Unified roadmap across core security, advanced architecture, and emergent security structures.

PhaseCore SecurityAdvanced ArchitectureEmergent SecurityTimeline
Phase 1 (MVP)SQLCipher, extension entitlements, content sanitization, domain exclusionKPC (protection classes), CSEB (event bus), CTA (trust arbiter)KSCI (supply chain integrity), TCV (temporal causality)Q3 2026
Phase 2 (Pro)Context TCC, Space boundaries, embedding encryption, differential privacyPSCS (privilege separation), SKV (sealed volumes), ASP (adaptive posture)IRD (inference residue defense), SBRC (blast radius containment)Q4 2026
Phase 3 (Enterprise)TEE integration, Context Lockdown, exfiltration detection, PQCKQR (query relay), DSPL (policy language)AGTR (graph topology resistance), CSELP (entropy leakage prevention)2027
Phase 4 (Standard)Context Security Framework — W3C/IETF standardization proposalFull integration + formal verificationRSIP (retroactive sanitization), CAV (amnesia verification)2027+
← Previous: Export & Import Next: Threat Model →
Was this page helpful?
Ask AI
Q

Ask anything about the documentation

Ctrl+K to open anytime