parlov-core 0.4.0

//! Shared types, error types, and oracle class definitions used across all parlov crates.
//!
//! This crate is the dependency root of the workspace — it carries no deps on other workspace
//! crates and is designed to compile fast. Everything in here is pure data: no I/O, no async,
//! no heavy dependencies.

#![deny(clippy::all)]
#![warn(clippy::pedantic)]
#![deny(missing_docs)]

mod exchange;
mod finding_id;
mod scoring;
mod serde_helpers;
mod signal;
mod technique;

pub use exchange::{DifferentialSet, ProbeExchange};
pub use finding_id::finding_id;
pub use scoring::{ScoringDimension, ScoringReason};
pub use signal::{ImpactClass, Signal, SignalKind};
pub use technique::{NormativeStrength, Technique, Vector};

use bytes::Bytes;
use http::{HeaderMap, Method, StatusCode};
use serde::{Deserialize, Serialize};
use serde_helpers::{
    bytes_serde, header_map_serde, method_serde, opt_bytes_serde, status_code_serde,
};

/// A single HTTP interaction: full response surface and wall-clock timing.
///
/// Captures everything needed for differential analysis — status, headers, body, and timing —
/// in one flat structure.
#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct ResponseSurface {
    /// HTTP status code returned by the server.
    #[serde(with = "status_code_serde")]
    pub status: StatusCode,
    /// Full response header map.
    #[serde(with = "header_map_serde")]
    pub headers: HeaderMap,
    /// Raw response body bytes, serialized as a base64-encoded byte sequence.
    #[serde(with = "bytes_serde")]
    pub body: Bytes,
    /// Wall-clock response time in nanoseconds, measured from first byte sent to last byte
    /// received.
    pub timing_ns: u64,
}

/// A single HTTP request to execute against a target.
///
/// The authorization context is expressed entirely through the `headers` field — set an
/// `Authorization` header for bearer tokens, API keys, or Basic auth. No special-case auth
/// fields.
#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct ProbeDefinition {
    /// Fully-qualified target URL including scheme, host, path, and any query parameters.
    pub url: String,
    /// HTTP method for the request.
    #[serde(with = "method_serde")]
    pub method: Method,
    /// Request headers, including any authorization context.
    #[serde(with = "header_map_serde")]
    pub headers: HeaderMap,
    /// Request body. `None` for GET, HEAD, DELETE; `Some` for POST, PATCH, PUT.
    #[serde(with = "opt_bytes_serde")]
    pub body: Option<Bytes>,
}

/// Paired response surfaces for differential analysis.
///
/// `baseline` holds responses for the control input (e.g. a known-existing resource ID).
/// `probe` holds responses for the variable input (e.g. a randomly generated nonexistent ID).
/// Multiple samples per side support statistical analysis for timing oracles.
///
/// # Deprecated
///
/// Use [`DifferentialSet`] instead. `DifferentialSet` pairs each response with the request that
/// produced it and carries [`Technique`] metadata end-to-end. This type will be removed in Pass 4.
#[deprecated(since = "0.4.0", note = "use DifferentialSet instead — see system-design.md")]
#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct ProbeSet {
    /// Responses for the known-valid / control input.
    pub baseline: Vec<ResponseSurface>,
    /// Responses for the unknown / suspect input.
    pub probe: Vec<ResponseSurface>,
}

/// The oracle class being probed.
///
/// Each variant corresponds to a distinct detection strategy and analysis pipeline.
#[derive(Debug, Clone, Copy, PartialEq, Eq, Hash, Serialize, Deserialize)]
#[non_exhaustive]
pub enum OracleClass {
    /// Status-code or body differential between an existing and nonexistent resource.
    Existence,
}

/// Confidence level of an oracle detection result.
#[derive(Debug, Clone, Copy, PartialEq, Eq, Hash, Serialize, Deserialize)]
pub enum OracleVerdict {
    /// Signal is unambiguous: differential is consistent, statistically significant, and matches
    /// a known oracle pattern.
    Confirmed,
    /// Signal is present and consistent with an oracle, but evidence is not conclusive (e.g.
    /// borderline p-value, single sample).
    Likely,
    /// Signal is present but too weak or inconsistent to classify.
    Inconclusive,
    /// No differential signal detected; the endpoint does not exhibit this oracle.
    NotPresent,
}

/// Severity of a confirmed or likely oracle.
///
/// `None` on an `OracleResult` when the verdict is `NotPresent` or `Inconclusive`.
#[derive(Debug, Clone, Copy, PartialEq, Eq, Hash, Serialize, Deserialize)]
pub enum Severity {
    /// Directly actionable: resource existence, valid credentials, or session state is leaked
    /// to unauthenticated or low-privilege callers.
    High,
    /// Leaks internal state but requires additional steps to exploit.
    Medium,
    /// Informational: leaks metadata that may assist further enumeration.
    Low,
}

/// The result of running an oracle analyzer against a differential set.
///
/// Carries the full signal chain that produced the verdict alongside technique context and
/// scoring breakdown. Status codes, header diffs, and flat evidence strings previously stored
/// in dedicated fields are now represented as typed [`Signal`] values in the `signals` vec.
#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct OracleResult {
    /// Which oracle class produced this result.
    pub class: OracleClass,
    /// Confidence verdict.
    pub verdict: OracleVerdict,
    /// Severity when the verdict is `Confirmed` or `Likely`; `None` when `NotPresent`.
    pub severity: Option<Severity>,
    /// Numeric confidence score (0-100). Determines verdict via threshold mapping.
    #[serde(default)]
    pub confidence: u8,
    /// Impact classification based on leak type. Determines severity when gated by confidence.
    #[serde(skip_serializing_if = "Option::is_none", default)]
    pub impact_class: Option<ImpactClass>,
    /// Breakdown of how confidence and impact were computed.
    #[serde(skip_serializing_if = "Vec::is_empty", default)]
    pub reasons: Vec<ScoringReason>,
    /// Typed signals extracted during differential analysis.
    #[serde(skip_serializing_if = "Vec::is_empty", default)]
    pub signals: Vec<Signal>,
    /// Machine-readable technique identifier, e.g. `"if-none-match"`.
    #[serde(skip_serializing_if = "Option::is_none", default)]
    pub technique_id: Option<String>,
    /// Detection vector used by the technique.
    #[serde(skip_serializing_if = "Option::is_none", default)]
    pub vector: Option<Vector>,
    /// RFC normative strength of the technique's expected differential.
    #[serde(skip_serializing_if = "Option::is_none", default)]
    pub normative_strength: Option<NormativeStrength>,
    /// Human-readable name for the detected pattern, e.g. `"Authorization-based differential"`.
    #[serde(skip_serializing_if = "Option::is_none", default)]
    pub label: Option<String>,
    /// What information the oracle leaks, e.g.
    /// `"Resource existence confirmed to low-privilege callers"`.
    #[serde(skip_serializing_if = "Option::is_none", default)]
    pub leaks: Option<String>,
    /// RFC section grounding the behavior, e.g. `"RFC 9110 \u{00a7}15.5.4"`.
    #[serde(skip_serializing_if = "Option::is_none", default)]
    pub rfc_basis: Option<String>,
}

impl OracleResult {
    /// Returns the evidence string from the primary `StatusCodeDiff` signal, if present.
    ///
    /// Falls back to the first signal of any kind, then `"—"` when no signals exist.
    #[must_use]
    pub fn primary_evidence(&self) -> &str {
        self.signals
            .iter()
            .find(|s| s.kind == SignalKind::StatusCodeDiff)
            .or_else(|| self.signals.first())
            .map_or("—", |s| s.evidence.as_str())
    }
}

/// Errors produced by parlov crates.
#[derive(Debug, thiserror::Error)]
#[non_exhaustive]
pub enum Error {
    /// HTTP-level error from the probe engine.
    #[error("http error: {0}")]
    Http(String),
    /// Analysis failed due to insufficient or malformed probe data.
    #[error("analysis error: {0}")]
    Analysis(String),
    /// Serialization or deserialization failure.
    #[error("serialization error: {0}")]
    Serialization(#[from] serde_json::Error),
}

#[cfg(test)]
mod tests {
    use super::*;

    fn confirmed_result_with_metadata() -> OracleResult {
        OracleResult {
            class: OracleClass::Existence,
            verdict: OracleVerdict::Confirmed,
            severity: Some(Severity::High),
            confidence: 0,
            impact_class: None,
            reasons: vec![],
            signals: vec![Signal {
                kind: SignalKind::StatusCodeDiff,
                evidence: "403 (baseline) vs 404 (probe)".into(),
                rfc_basis: None,
            }],
            technique_id: None,
            vector: None,
            normative_strength: None,
            label: Some("Authorization-based differential".into()),
            leaks: Some("Resource existence confirmed to low-privilege callers".into()),
            rfc_basis: Some("RFC 9110 \u{00a7}15.5.4".into()),
        }
    }

    fn not_present_result() -> OracleResult {
        OracleResult {
            class: OracleClass::Existence,
            verdict: OracleVerdict::NotPresent,
            severity: None,
            confidence: 0,
            impact_class: None,
            reasons: vec![],
            signals: vec![Signal {
                kind: SignalKind::StatusCodeDiff,
                evidence: "404 (baseline) vs 404 (probe)".into(),
                rfc_basis: None,
            }],
            technique_id: None,
            vector: None,
            normative_strength: None,
            label: None,
            leaks: None,
            rfc_basis: None,
        }
    }

    #[test]
    fn serialize_confirmed_includes_metadata_fields() {
        let result = confirmed_result_with_metadata();
        let json = serde_json::to_value(&result).expect("serialization failed");
        assert_eq!(json["label"], "Authorization-based differential");
        assert_eq!(json["leaks"], "Resource existence confirmed to low-privilege callers");
        assert_eq!(json["rfc_basis"], "RFC 9110 \u{00a7}15.5.4");
    }

    #[test]
    fn serialize_not_present_omits_none_metadata() {
        let result = not_present_result();
        let json = serde_json::to_value(&result).expect("serialization failed");
        assert!(!json.as_object().expect("expected object").contains_key("label"));
        assert!(!json.as_object().expect("expected object").contains_key("leaks"));
        assert!(!json.as_object().expect("expected object").contains_key("rfc_basis"));
    }

    #[test]
    fn roundtrip_confirmed_preserves_metadata() {
        let original = confirmed_result_with_metadata();
        let json = serde_json::to_string(&original).expect("serialization failed");
        let deserialized: OracleResult =
            serde_json::from_str(&json).expect("deserialization failed");
        assert_eq!(deserialized.label, original.label);
        assert_eq!(deserialized.leaks, original.leaks);
        assert_eq!(deserialized.rfc_basis, original.rfc_basis);
    }

    #[test]
    fn roundtrip_not_present_preserves_none_metadata() {
        let original = not_present_result();
        let json = serde_json::to_string(&original).expect("serialization failed");
        let deserialized: OracleResult =
            serde_json::from_str(&json).expect("deserialization failed");
        assert_eq!(deserialized.label, None);
        assert_eq!(deserialized.leaks, None);
        assert_eq!(deserialized.rfc_basis, None);
    }

    #[test]
    fn deserialize_minimal_json_defaults_to_none() {
        let minimal = r#"{
            "class": "Existence",
            "verdict": "Confirmed",
            "severity": "High"
        }"#;
        let result: OracleResult =
            serde_json::from_str(minimal).expect("deserialization failed");
        assert_eq!(result.label, None);
        assert_eq!(result.leaks, None);
        assert_eq!(result.rfc_basis, None);
        assert!(result.signals.is_empty());
        assert_eq!(result.technique_id, None);
        assert_eq!(result.confidence, 0);
        assert_eq!(result.impact_class, None);
        assert!(result.reasons.is_empty());
    }

    #[test]
    fn oracle_result_with_technique_context_serializes() {
        let result = OracleResult {
            class: OracleClass::Existence,
            verdict: OracleVerdict::Confirmed,
            severity: Some(Severity::High),
            confidence: 0,
            impact_class: None,
            reasons: vec![],
            signals: vec![Signal {
                kind: SignalKind::StatusCodeDiff,
                evidence: "304 vs 404".into(),
                rfc_basis: Some("RFC 9110 \u{00a7}13.1.2".into()),
            }],
            technique_id: Some("if-none-match".into()),
            vector: Some(Vector::CacheProbing),
            normative_strength: Some(NormativeStrength::Must),
            label: None,
            leaks: None,
            rfc_basis: None,
        };
        let json = serde_json::to_value(&result).expect("serialization failed");
        assert_eq!(json["technique_id"], "if-none-match");
        assert_eq!(json["vector"], "CacheProbing");
        assert_eq!(json["normative_strength"], "Must");
        assert_eq!(json["signals"][0]["kind"], "StatusCodeDiff");
        assert_eq!(json["signals"][0]["evidence"], "304 vs 404");
        assert_eq!(json["signals"][0]["rfc_basis"], "RFC 9110 \u{00a7}13.1.2");
    }

    #[test]
    fn oracle_result_roundtrip_with_technique_context() {
        let original = OracleResult {
            class: OracleClass::Existence,
            verdict: OracleVerdict::Likely,
            severity: Some(Severity::Medium),
            confidence: 0,
            impact_class: None,
            reasons: vec![],
            signals: vec![Signal {
                kind: SignalKind::HeaderPresence,
                evidence: "ETag present in baseline, absent in probe".into(),
                rfc_basis: None,
            }],
            technique_id: Some("get-200-404".into()),
            vector: Some(Vector::StatusCodeDiff),
            normative_strength: Some(NormativeStrength::Should),
            label: Some("Status code differential".into()),
            leaks: Some("Resource existence".into()),
            rfc_basis: Some("RFC 9110 \u{00a7}15.5.5".into()),
        };
        let json = serde_json::to_string(&original).expect("serialization failed");
        let back: OracleResult = serde_json::from_str(&json).expect("deserialization failed");
        assert_eq!(back.technique_id, original.technique_id);
        assert_eq!(back.vector, original.vector);
        assert_eq!(back.normative_strength, original.normative_strength);
        assert_eq!(back.signals.len(), 1);
        assert_eq!(back.signals[0].kind, SignalKind::HeaderPresence);
    }

    #[test]
    fn primary_evidence_returns_status_code_diff() {
        let result = confirmed_result_with_metadata();
        assert_eq!(result.primary_evidence(), "403 (baseline) vs 404 (probe)");
    }

    #[test]
    fn primary_evidence_falls_back_to_first_signal() {
        let result = OracleResult {
            class: OracleClass::Existence,
            verdict: OracleVerdict::Confirmed,
            severity: Some(Severity::Medium),
            confidence: 0,
            impact_class: None,
            reasons: vec![],
            signals: vec![Signal {
                kind: SignalKind::HeaderPresence,
                evidence: "etag present in baseline".into(),
                rfc_basis: None,
            }],
            technique_id: None,
            vector: None,
            normative_strength: None,
            label: None,
            leaks: None,
            rfc_basis: None,
        };
        assert_eq!(result.primary_evidence(), "etag present in baseline");
    }

    #[test]
    fn primary_evidence_returns_dash_when_empty() {
        let result = not_present_result();
        let mut empty = result;
        empty.signals.clear();
        assert_eq!(empty.primary_evidence(), "\u{2014}");
    }

    #[test]
    fn signal_kind_copy_and_eq() {
        let a = SignalKind::StatusCodeDiff;
        let b = a;
        assert_eq!(a, b);
    }

    #[test]
    fn vector_copy_and_eq() {
        let a = Vector::CacheProbing;
        let b = a;
        assert_eq!(a, b);
    }

    #[test]
    fn normative_strength_copy_and_eq() {
        let a = NormativeStrength::Must;
        let b = a;
        assert_eq!(a, b);
        assert_ne!(a, NormativeStrength::May);
    }

    #[test]
    fn technique_clone() {
        let t = Technique {
            id: "test",
            name: "Test technique",
            oracle_class: OracleClass::Existence,
            vector: Vector::StatusCodeDiff,
            strength: NormativeStrength::Must,
        };
        let t2 = t.clone();
        assert_eq!(t2.id, "test");
        assert_eq!(t2.vector, Vector::StatusCodeDiff);
    }

    #[test]
    fn probe_exchange_pairs_request_and_response() {
        let exchange = ProbeExchange {
            request: ProbeDefinition {
                url: "https://example.com/resource/1".into(),
                method: http::Method::GET,
                headers: HeaderMap::new(),
                body: None,
            },
            response: ResponseSurface {
                status: http::StatusCode::OK,
                headers: HeaderMap::new(),
                body: Bytes::new(),
                timing_ns: 1_000_000,
            },
        };
        assert_eq!(exchange.request.url, "https://example.com/resource/1");
        assert_eq!(exchange.response.status, http::StatusCode::OK);
    }

    #[test]
    fn differential_set_carries_technique() {
        let technique = Technique {
            id: "get-200-404",
            name: "GET 200/404",
            oracle_class: OracleClass::Existence,
            vector: Vector::StatusCodeDiff,
            strength: NormativeStrength::Must,
        };
        let ds = DifferentialSet {
            baseline: vec![],
            probe: vec![],
            technique,
        };
        assert_eq!(ds.technique.id, "get-200-404");
        assert_eq!(ds.technique.strength, NormativeStrength::Must);
    }

    #[test]
    fn technique_without_target_signals_constructs() {
        let t = Technique {
            id: "range-416",
            name: "Range 416/404",
            oracle_class: OracleClass::Existence,
            vector: Vector::CacheProbing,
            strength: NormativeStrength::Should,
        };
        let t2 = t.clone();
        assert_eq!(t2.id, "range-416");
        assert_eq!(t2.strength, NormativeStrength::Should);
        assert_eq!(t2.oracle_class, OracleClass::Existence);
    }

    #[test]
    fn impact_class_copy_and_eq() {
        let a = ImpactClass::High;
        let b = a;
        assert_eq!(a, b);
        assert_ne!(a, ImpactClass::Low);
    }

    #[test]
    fn impact_class_serialize_roundtrip() {
        let json = serde_json::to_string(&ImpactClass::Medium).expect("serialization failed");
        let back: ImpactClass = serde_json::from_str(&json).expect("deserialization failed");
        assert_eq!(back, ImpactClass::Medium);
    }

    #[test]
    fn scoring_dimension_copy_and_eq() {
        let a = ScoringDimension::Confidence;
        let b = a;
        assert_eq!(a, b);
        assert_ne!(a, ScoringDimension::Impact);
    }

    #[test]
    fn scoring_reason_serialize_roundtrip() {
        let reason = ScoringReason {
            description: "Status differential 416 vs 404".into(),
            points: 75,
            dimension: ScoringDimension::Confidence,
        };
        let json = serde_json::to_string(&reason).expect("serialization failed");
        let back: ScoringReason = serde_json::from_str(&json).expect("deserialization failed");
        assert_eq!(back.description, "Status differential 416 vs 404");
        assert_eq!(back.points, 75);
        assert_eq!(back.dimension, ScoringDimension::Confidence);
    }

    #[test]
    fn scoring_reason_negative_points() {
        let reason = ScoringReason {
            description: "Inconsistent across samples".into(),
            points: -10,
            dimension: ScoringDimension::Confidence,
        };
        let json = serde_json::to_string(&reason).expect("serialization failed");
        let back: ScoringReason = serde_json::from_str(&json).expect("deserialization failed");
        assert_eq!(back.points, -10);
    }

    #[test]
    fn oracle_result_with_confidence_and_impact_serializes() {
        let result = OracleResult {
            class: OracleClass::Existence,
            verdict: OracleVerdict::Confirmed,
            severity: Some(Severity::High),
            confidence: 88,
            impact_class: Some(ImpactClass::High),
            reasons: vec![
                ScoringReason {
                    description: "Status differential 416 vs 404".into(),
                    points: 75,
                    dimension: ScoringDimension::Confidence,
                },
                ScoringReason {
                    description: "Content-Range reveals exact size".into(),
                    points: 12,
                    dimension: ScoringDimension::Impact,
                },
            ],
            signals: vec![],
            technique_id: None,
            vector: None,
            normative_strength: None,
            label: None,
            leaks: None,
            rfc_basis: None,
        };
        let json = serde_json::to_value(&result).expect("serialization failed");
        assert_eq!(json["confidence"], 88);
        assert_eq!(json["impact_class"], "High");
        assert_eq!(json["reasons"].as_array().expect("expected array").len(), 2);
        assert_eq!(json["reasons"][0]["points"], 75);
        assert_eq!(json["reasons"][1]["dimension"], "Impact");
    }

    #[test]
    fn oracle_result_zero_confidence_omits_impact_and_reasons() {
        let result = OracleResult {
            class: OracleClass::Existence,
            verdict: OracleVerdict::NotPresent,
            severity: None,
            confidence: 0,
            impact_class: None,
            reasons: vec![],
            signals: vec![],
            technique_id: None,
            vector: None,
            normative_strength: None,
            label: None,
            leaks: None,
            rfc_basis: None,
        };
        let json = serde_json::to_value(&result).expect("serialization failed");
        let obj = json.as_object().expect("expected object");
        assert!(!obj.contains_key("impact_class"));
        assert!(!obj.contains_key("reasons"));
    }
}