secretsniff-core 0.1.0

//! Pure-Rust core for `secretsniff`. Source-code secret scanner.
//!
//! Two layers:
//!
//! - **Regex detectors** for known formats (AWS keys, GitHub PATs, etc.).
//!   These are fast, low-false-positive when patterns match exactly.
//! - **High-entropy filter** that flags any base64/hex-ish substring of
//!   length ≥ `min_entropy_length` whose Shannon entropy meets a
//!   threshold. Catches one-off secrets that don't fit a known format.

#![deny(unsafe_code)]
#![warn(missing_docs)]
#![warn(rust_2018_idioms)]

use rayon::prelude::*;
use regex::Regex;
use serde::{Deserialize, Serialize};
use thiserror::Error;

/// Crate-wide result alias.
pub type Result<T> = std::result::Result<T, ScannerError>;

/// All errors surfaced by `secretsniff-core`.
#[derive(Error, Debug)]
pub enum ScannerError {
    /// A regex failed to compile.
    #[error("regex error: {0}")]
    Regex(#[from] regex::Error),
    /// Caller supplied an invalid configuration.
    #[error("invalid config: {0}")]
    InvalidConfig(String),
}

/// Scanner configuration.
#[derive(Debug, Clone, PartialEq, Serialize, Deserialize)]
pub struct ScannerConfig {
    /// Shannon-entropy threshold (bits per char) for the high-entropy
    /// fallback rule. 4.5 is a reasonable starting point.
    pub min_entropy: f32,
    /// Minimum length (in characters) for the high-entropy rule.
    pub min_entropy_length: usize,
    /// If false, skip the high-entropy rule entirely.
    pub include_high_entropy: bool,
}

impl Default for ScannerConfig {
    fn default() -> Self {
        Self {
            min_entropy: 4.5,
            min_entropy_length: 32,
            include_high_entropy: true,
        }
    }
}

/// One finding.
#[derive(Debug, Clone, PartialEq, Serialize, Deserialize)]
pub struct Finding {
    /// Detector name (e.g. `AWS_ACCESS_KEY`).
    pub kind: String,
    /// 1-indexed line number.
    pub line: usize,
    /// 1-indexed byte offset within the line.
    pub column: usize,
    /// Byte offset of the match start in the source.
    pub start: usize,
    /// Byte offset (exclusive) of the match end.
    pub end: usize,
    /// The matched substring.
    pub matched: String,
    /// Shannon entropy in bits/char of the matched string.
    pub entropy: f32,
}

/// Compiled scanner.
pub struct Scanner {
    cfg: ScannerConfig,
    rules: Vec<(&'static str, Regex)>,
    high_entropy_re: Regex,
}

const RULES: &[(&str, &str)] = &[
    ("AWS_ACCESS_KEY", r"\bAKIA[0-9A-Z]{16}\b"),
    // GitHub token formats: ghp_ (PAT), gho_ (OAuth), ghu_ (user-to-server),
    // ghr_ (refresh), ghs_ (server-to-server).
    ("GITHUB_TOKEN", r"\bgh[pours]_[A-Za-z0-9]{36,}\b"),
    // Slack tokens
    ("SLACK_TOKEN", r"\bxox[baprs]-[A-Za-z0-9-]{10,}\b"),
    // Stripe keys
    (
        "STRIPE_KEY",
        r"\b(?:sk|pk|rk)_(?:live|test)_[A-Za-z0-9]{20,}\b",
    ),
    // JWT (three url-safe-base64 segments)
    (
        "JWT",
        r"\beyJ[A-Za-z0-9_-]+\.eyJ[A-Za-z0-9_-]+\.[A-Za-z0-9_-]+\b",
    ),
    // PEM-style markers
    ("RSA_PRIVATE_KEY", r"-----BEGIN RSA PRIVATE KEY-----"),
    ("SSH_PRIVATE_KEY", r"-----BEGIN OPENSSH PRIVATE KEY-----"),
    // Generic api_key = "..." assignments. Captures the value via a group.
    (
        "GENERIC_API_KEY",
        r#"(?i)\bapi[_-]?key\s*[=:]\s*['"]([A-Za-z0-9_\-=]{16,})['"]"#,
    ),
];

impl Scanner {
    /// Build a scanner with the default config (all rules enabled).
    pub fn new() -> Self {
        Self::with_config(ScannerConfig::default()).expect("default config compiles")
    }

    /// Build a scanner with a custom config.
    pub fn with_config(cfg: ScannerConfig) -> Result<Self> {
        if cfg.min_entropy < 0.0 || cfg.min_entropy > 8.0 {
            return Err(ScannerError::InvalidConfig(format!(
                "min_entropy out of range [0, 8]: {}",
                cfg.min_entropy
            )));
        }
        let rules: Vec<(&'static str, Regex)> = RULES
            .iter()
            .map(|(k, p)| Regex::new(p).map(|r| (*k, r)))
            .collect::<std::result::Result<_, _>>()?;
        // High-entropy candidate match: substrings of length >= min that look
        // like base64/hex/url-safe-base64.
        let pat = format!(r"[A-Za-z0-9+/=_\-]{{{},}}", cfg.min_entropy_length);
        let high_entropy_re = Regex::new(&pat)?;
        Ok(Self {
            cfg,
            rules,
            high_entropy_re,
        })
    }

    /// Scan `source`, returning findings in source order.
    pub fn scan(&self, source: &str) -> Vec<Finding> {
        let mut findings: Vec<Finding> = Vec::new();
        let mut covered: Vec<(usize, usize)> = Vec::new();

        // Built-in rules.
        for (kind, regex) in &self.rules {
            for m in regex.find_iter(source) {
                let entropy = shannon_entropy(m.as_str());
                let (line, column) = line_col(source, m.start());
                findings.push(Finding {
                    kind: (*kind).to_string(),
                    line,
                    column,
                    start: m.start(),
                    end: m.end(),
                    matched: m.as_str().to_string(),
                    entropy,
                });
                covered.push((m.start(), m.end()));
            }
        }

        // High-entropy fallback. Skip ranges already covered.
        if self.cfg.include_high_entropy {
            for m in self.high_entropy_re.find_iter(source) {
                if overlaps(&covered, m.start(), m.end()) {
                    continue;
                }
                let entropy = shannon_entropy(m.as_str());
                if entropy < self.cfg.min_entropy {
                    continue;
                }
                let (line, column) = line_col(source, m.start());
                findings.push(Finding {
                    kind: "HIGH_ENTROPY".to_string(),
                    line,
                    column,
                    start: m.start(),
                    end: m.end(),
                    matched: m.as_str().to_string(),
                    entropy,
                });
            }
        }

        findings.sort_by_key(|f| f.start);
        findings
    }

    /// Bulk scan. With `parallel = true`, distributes across rayon's pool.
    pub fn scan_many(&self, sources: &[&str], parallel: bool) -> Vec<Vec<Finding>> {
        if parallel {
            sources.par_iter().map(|s| self.scan(s)).collect()
        } else {
            sources.iter().map(|s| self.scan(s)).collect()
        }
    }
}

impl Default for Scanner {
    fn default() -> Self {
        Self::new()
    }
}

fn shannon_entropy(s: &str) -> f32 {
    if s.is_empty() {
        return 0.0;
    }
    let mut counts = [0u32; 256];
    let mut n = 0u32;
    for &b in s.as_bytes() {
        counts[b as usize] += 1;
        n += 1;
    }
    let mut e = 0.0_f32;
    let n_f = n as f32;
    for &c in &counts {
        if c == 0 {
            continue;
        }
        let p = c as f32 / n_f;
        e -= p * p.log2();
    }
    e
}

fn line_col(source: &str, byte_offset: usize) -> (usize, usize) {
    let mut line = 1usize;
    let mut last_newline = 0usize;
    for (i, b) in source.as_bytes().iter().take(byte_offset).enumerate() {
        if *b == b'\n' {
            line += 1;
            last_newline = i + 1;
        }
    }
    (line, byte_offset - last_newline + 1)
}

fn overlaps(ranges: &[(usize, usize)], start: usize, end: usize) -> bool {
    ranges.iter().any(|&(s, e)| start < e && end > s)
}

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

    #[test]
    fn aws_key_detected() {
        let s = Scanner::new();
        let r = s.scan("aws = AKIAIOSFODNN7EXAMPLE\n");
        assert_eq!(r.len(), 1);
        assert_eq!(r[0].kind, "AWS_ACCESS_KEY");
        assert_eq!(r[0].line, 1);
    }

    #[test]
    fn github_token_detected() {
        let s = Scanner::new();
        let token = "ghp_abcdefghijklmnopqrstuvwxyz0123456789";
        let r = s.scan(&format!("token = {token}\n"));
        assert!(r.iter().any(|f| f.kind == "GITHUB_TOKEN"));
    }

    #[test]
    fn slack_token_detected() {
        let s = Scanner::new();
        let r = s.scan("slack = xoxb-1234567890-abcdef\n");
        assert!(r.iter().any(|f| f.kind == "SLACK_TOKEN"));
    }

    #[test]
    fn stripe_key_detected() {
        let s = Scanner::new();
        let r = s.scan("STRIPE = sk_live_abcdefghij1234567890\n");
        assert!(r.iter().any(|f| f.kind == "STRIPE_KEY"));
    }

    #[test]
    fn jwt_detected() {
        let s = Scanner::new();
        let jwt = "eyJhbGciOiJIUzI1NiJ9.eyJzdWIiOiJ1MSJ9.signature_part_long_enough";
        let r = s.scan(&format!("auth = '{jwt}'"));
        assert!(r.iter().any(|f| f.kind == "JWT"));
    }

    #[test]
    fn rsa_marker_detected() {
        let s = Scanner::new();
        let r = s.scan("-----BEGIN RSA PRIVATE KEY-----\nMII...\n");
        assert!(r.iter().any(|f| f.kind == "RSA_PRIVATE_KEY"));
    }

    #[test]
    fn ssh_marker_detected() {
        let s = Scanner::new();
        let r = s.scan("-----BEGIN OPENSSH PRIVATE KEY-----\nb3Bl...\n");
        assert!(r.iter().any(|f| f.kind == "SSH_PRIVATE_KEY"));
    }

    #[test]
    fn generic_api_key_assignment_detected() {
        let s = Scanner::new();
        let r = s.scan(r#"api_key = "abcdefghijklmnopqrst""#);
        assert!(r.iter().any(|f| f.kind == "GENERIC_API_KEY"));
    }

    #[test]
    fn high_entropy_detected() {
        let s = Scanner::new();
        // 32 chars of varied base64-ish content; entropy is high.
        let blob = "K3s9Q2pXq9ZTm4Lp2Vw7Yc1RnFb5Xh6N";
        let r = s.scan(&format!("token = '{blob}'"));
        assert!(r.iter().any(|f| f.kind == "HIGH_ENTROPY"));
    }

    #[test]
    fn low_entropy_skipped() {
        let s = Scanner::new();
        // 32 chars of a low-entropy string ('aaaa...').
        let blob = "aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa";
        let r = s.scan(&format!("v = '{blob}'"));
        assert!(!r.iter().any(|f| f.kind == "HIGH_ENTROPY"));
    }

    #[test]
    fn no_finding_on_clean_source() {
        let s = Scanner::new();
        let r = s.scan("def add(a: int, b: int) -> int:\n    return a + b\n");
        assert!(r.is_empty());
    }

    #[test]
    fn line_column_correct_on_multiline() {
        let s = Scanner::new();
        let src = "line1\nline2 AKIAIOSFODNN7EXAMPLE\nline3\n";
        let r = s.scan(src);
        assert_eq!(r.len(), 1);
        assert_eq!(r[0].line, 2);
        // "line2 " is 6 bytes; AKIA starts at column 7.
        assert_eq!(r[0].column, 7);
    }

    #[test]
    fn findings_sorted_by_position() {
        let s = Scanner::new();
        let src = "ghp_abcdefghijklmnopqrstuvwxyz0123456789 then AKIAIOSFODNN7EXAMPLE";
        let r = s.scan(src);
        assert!(r.len() >= 2);
        for w in r.windows(2) {
            assert!(w[0].start <= w[1].start);
        }
    }

    #[test]
    fn high_entropy_does_not_double_up_on_known_pattern() {
        let s = Scanner::new();
        // ghp_ token would also score as high-entropy. We should only get
        // GITHUB_TOKEN, not also HIGH_ENTROPY for the same span.
        let token = "ghp_abcdefghijklmnopqrstuvwxyz0123456789";
        let r = s.scan(&format!("t = '{token}'"));
        let kinds: Vec<&str> = r.iter().map(|f| f.kind.as_str()).collect();
        assert!(kinds.contains(&"GITHUB_TOKEN"));
        // HIGH_ENTROPY may still appear if it matches a different range, but
        // not over the GITHUB_TOKEN span.
        for f in &r {
            if f.kind == "HIGH_ENTROPY" {
                let token_start = src_pos(&format!("t = '{token}'"), token);
                let token_end = token_start + token.len();
                assert!(
                    !(f.start >= token_start && f.end <= token_end),
                    "HIGH_ENTROPY overlaps GITHUB_TOKEN at {}..{}",
                    f.start,
                    f.end
                );
            }
        }
    }

    fn src_pos(haystack: &str, needle: &str) -> usize {
        haystack.find(needle).unwrap()
    }

    #[test]
    fn invalid_entropy_threshold_rejected() {
        let cfg = ScannerConfig {
            min_entropy: 100.0,
            ..Default::default()
        };
        assert!(Scanner::with_config(cfg).is_err());
    }

    #[test]
    fn high_entropy_can_be_disabled() {
        let cfg = ScannerConfig {
            include_high_entropy: false,
            ..Default::default()
        };
        let s = Scanner::with_config(cfg).unwrap();
        let blob = "K3s9Q2pXq9ZTm4Lp2Vw7Yc1RnFb5Xh6N";
        let r = s.scan(&format!("token = '{blob}'"));
        assert!(!r.iter().any(|f| f.kind == "HIGH_ENTROPY"));
    }

    #[test]
    fn scan_many_serial_and_parallel_match() {
        let s = Scanner::new();
        let sources: Vec<&str> = vec!["aws = AKIAIOSFODNN7EXAMPLE", "no secret here"];
        let a = s.scan_many(&sources, false);
        let b = s.scan_many(&sources, true);
        assert_eq!(a, b);
        assert_eq!(a[0].len(), 1);
        assert_eq!(a[1].len(), 0);
    }

    #[test]
    fn shannon_entropy_zero_for_constant_string() {
        assert_eq!(shannon_entropy("aaaa"), 0.0);
    }

    #[test]
    fn shannon_entropy_max_for_equal_distribution() {
        // 4 distinct chars equally distributed -> 2 bits/char.
        let e = shannon_entropy("abcdabcd");
        assert!((e - 2.0).abs() < 1e-4);
    }
}