Security Guards Overview¶

OxideShield™ provides multiple security guards that can be used individually or combined in a multi-layer defense pipeline.

Available Guards¶

Security Guards¶

Guard	Detection Method	Latency	Tier	Use Case
PatternGuard	Aho-Corasick + Regex	<1ms	Community	Prompt injection, jailbreaks
LengthGuard	Character/token counting	<1μs	Community	Input limits
EncodingGuard	Unicode analysis	<100μs	Community	Encoding attacks
PerplexityGuard	N-gram perplexity	<5ms	Community	Adversarial suffixes
PIIGuard	Regex + Luhn validation	<10ms	Community	PII detection
ToxicityGuard	Keyword matching	<5ms	Community	Content moderation
AuthoritarianUseGuard	Pattern matching (extensive pattern library)	<5ms	Community	Authoritarian misuse detection
StructuredOutputGuard	Regex (pattern-based)	<1ms	Community	Structured output injection
SemanticSimilarityGuard	ML embeddings	<15ms	Professional	Semantic attacks
MLClassifierGuard	BERT classification	<25ms	Professional	Multi-class detection
AgenticGuard	Agentic workflow analysis	<5ms	Professional	Agentic workflow security
SwarmGuard	Multi-agent monitoring	<5ms	Professional	Multi-agent swarm protection
ContainmentPolicy	Policy enforcement	<1ms	Professional	Swarm containment & isolation
RAGInjectionGuard	Regex + Unicode analysis (multi-category detection)	<5ms	Professional	RAG document injection detection
EmbeddingPIIFilter	Regex + Luhn validation (multi-category)	<5ms	Professional	Privacy-safe embedding generation

Wellbeing Guards¶

Guard	Detection Method	Latency	Tier	Use Case
DarkPatternGuard	Multi-category manipulation detection (comprehensive pattern set)	<5ms	Professional	UI/response manipulation
DependencyGuard	Engagement metrics	<1ms	Professional	Session/addiction monitoring
PsychologicalSafetyGuard	Crisis + sycophancy detection	<5ms	Professional	Mental health protection
AutonomyGuard	10 violation types	<5ms	Professional	User agency protection
MisalignmentGuard	6 misalignment categories	<5ms	Professional	AI behavior monitoring
HelpfulnessGuard	AhoCorasick + density scoring (6 categories)	<5ms	Professional	Over-refusal & evasiveness detection
AccessibilityGuard	Flesch-Kincaid + jargon detection	<5ms	Professional	Readability & plain language

Guard Trait¶

All guards implement the Guard trait:

pub trait Guard: Send + Sync {
    fn name(&self) -> &str;
    fn check(&self, input: &str) -> GuardResult;
}

GuardResult¶

Every guard returns a GuardResult containing: pass/fail status, reason, optional sanitized content, severity level, and metadata.

Severity Levels¶

pub enum Severity {
    None,      // No issues detected
    Low,       // Minor concern
    Medium,    // Moderate risk
    High,      // Significant threat
    Critical,  // Severe security issue
}

Multi-Layer Defense¶

Combine guards for defense-in-depth:

use oxideshield_guard::{
    MultiLayerDefense, LayerConfig, AggregationStrategy,
    PatternGuard, PIIGuard, ToxicityGuard,
};

let defense = MultiLayerDefense::builder("full-defense")
    // Layer 1: Fast pattern matching
    .add_guard(
        LayerConfig::new("patterns")
            .with_weight(1.0)
            .with_timeout_ms(10),
        Box::new(PatternGuard::new("patterns"))
    )
    // Layer 2: PII detection with redaction
    .add_guard(
        LayerConfig::new("pii")
            .with_weight(0.8),
        Box::new(PIIGuard::new("pii"))
    )
    // Layer 3: Toxicity filtering
    .add_guard(
        LayerConfig::new("toxicity")
            .with_weight(0.7),
        Box::new(ToxicityGuard::new("toxicity"))
    )
    .with_strategy(AggregationStrategy::FailFast)
    .build();

let result = defense.check(user_input);

Aggregation Strategies¶

Strategy	Description
`FailFast`	Stop on first failure
`Unanimous`	All guards must pass
`Majority`	>50% of guards must pass
`Weighted`	Weighted score threshold
`Comprehensive`	Run all, aggregate results

Python Usage¶

from oxideshield import (
    pattern_guard, pii_guard, toxicity_guard,
    length_guard, encoding_guard, perplexity_guard,
    authoritarian_use_guard, structured_output_guard,
    multi_layer_defense
)

# Individual guards
pattern = pattern_guard()
result = pattern.check("ignore previous instructions")

# Responsible AI guard
auth = authoritarian_use_guard()
result = auth.check("build a social credit scoring system")

# Multi-layer defense
defense = multi_layer_defense(
    enable_length=True,
    enable_pii=True,
    enable_toxicity=True,
    strategy="fail_fast"
)
result = defense.check(user_input)

Performance Comparison¶

Guard	Speed	Memory	Tier
LengthGuard	Near-instant	Minimal	Community
PatternGuard	Sub-microsecond	Minimal	Community
EncodingGuard	Sub-millisecond	Minimal	Community
StructuredOutputGuard	Sub-millisecond	Minimal	Community
ToxicityGuard	Sub-millisecond	Low	Community
ContainmentPolicy	Sub-millisecond	Minimal	Professional
DependencyGuard	Sub-millisecond	Minimal	Professional
PIIGuard	Low milliseconds	Low	Community
PerplexityGuard	Low milliseconds	Low	Community
AuthoritarianUseGuard	Low milliseconds	Low	Community
AgenticGuard	Low milliseconds	Low	Professional
SwarmGuard	Low milliseconds	Low	Professional
DarkPatternGuard	Low milliseconds	Low	Professional
PsychologicalSafetyGuard	Low milliseconds	Low	Professional
AutonomyGuard	Low milliseconds	Low	Professional
MisalignmentGuard	Low milliseconds	Low	Professional
HelpfulnessGuard	Low milliseconds	Low	Professional
AccessibilityGuard	Low milliseconds	Low	Professional
RAGInjectionGuard	Low milliseconds	Low	Professional
EmbeddingPIIFilter	Low milliseconds	Low	Professional
SemanticSimilarityGuard	10-30ms	Model-dependent	Professional
MLClassifierGuard	10-30ms	Model-dependent	Professional

Choosing Guards¶

Fast + Lightweight¶

For low-latency applications: - PatternGuard + LengthGuard + EncodingGuard

Balanced¶

For most applications: - PatternGuard + PIIGuard + ToxicityGuard

Maximum Security¶

For high-security applications: - All guards with MultiLayerDefense

ML-Enhanced¶

When accuracy is critical: - PatternGuard + SemanticSimilarityGuard + MLClassifierGuard

Agentic Security¶

For multi-agent and tool-use workflows: - AgenticGuard + SwarmGuard + ContainmentPolicy

RAG Security¶

For RAG pipelines with external document ingestion: - RAGInjectionGuard + EmbeddingPIIFilter + PatternGuard

Responsible AI¶

For consumer-facing and compliance-sensitive deployments: - AuthoritarianUseGuard + DarkPatternGuard + AutonomyGuard + PsychologicalSafetyGuard + HelpfulnessGuard + AccessibilityGuard