captchaforge 0.2.7

//! Math-captcha solver.
//!
//! Many low-end captcha widgets (especially WordPress plugins, basic
//! contact forms, and homegrown anti-spam layers) ask the user to
//! solve a simple arithmetic question:
//!
//! - "What is 3 + 5?"
//! - "7 × 2 = ?"
//! - "Type the result of 12 - 4"
//! - "What is two plus four?"  (number words)
//!
//! The challenge is plain text in the DOM and the answer goes in a
//! visible `<input>`. captchaforge can solve these without VLM, OCR,
//! or third-party calls — just parse, evaluate, type, submit. Free,
//! ~50ms wall-clock.
//!
//! Supported expression shapes:
//!
//! - `N OP N` where OP ∈ `+ - * × x ÷ /` and N is a digit string or
//!   the spelled-out word for 0–20.
//! - Order doesn't matter: "What is 3 + 5", "3 plus 5 equals what",
//!   "Solve: 3 + 5 =" all parse to the same expression.
//! - Both decimal and word number representations.
//!
//! Out of scope (returns failure):
//!
//! - Multi-operator expressions ("3 + 5 - 2") — would need a full
//!   parser; rare in math captchas.
//! - Algebraic ("if x + 3 = 7, what is x?") — different solver.
//! - Word problems ("Tom has 3 apples...") — VLM territory.

use super::*;
use crate::captcha_detect::DetectedCaptcha;
use std::time::Instant;

/// Selectors we'll search for the challenge text. Ordered most to
/// least specific — first hit wins.
const QUESTION_SELECTORS: &[&str] = &[
    "[id*='math-captcha'] label",
    "[class*='math-captcha'] label",
    "[id*='math_captcha'] label",
    "[class*='math_captcha'] label",
    ".captcha-question",
    ".captcha-prompt",
    "label[for*='captcha']",
    "label[for*='math']",
    "[data-captcha-question]",
    ".g-recaptcha-question",
];

/// Selectors for the answer input.
const ANSWER_SELECTORS: &[&str] = &[
    "input[name*='math-captcha']",
    "input[name*='math_captcha']",
    "input[name='captcha']",
    "input[name='captcha_answer']",
    "input[id*='captcha-answer']",
    "input[name*='quiz']",
    "input[type='text'][placeholder*='answer']",
    "input[type='number'][name*='captcha']",
];

/// JS that scans the page for a likely math-captcha question + input
/// and returns both. Single round-trip.
pub(crate) const PROBE_JS: &str = r#"
(() => {
    const QUESTION_SELECTORS = [
        "[id*='math-captcha'] label",
        "[class*='math-captcha'] label",
        "[id*='math_captcha'] label",
        "[class*='math_captcha'] label",
        ".captcha-question",
        ".captcha-prompt",
        "label[for*='captcha']",
        "label[for*='math']",
        "[data-captcha-question]",
        ".g-recaptcha-question"
    ];
    const ANSWER_SELECTORS = [
        "input[name*='math-captcha']",
        "input[name*='math_captcha']",
        "input[name='captcha']",
        "input[name='captcha_answer']",
        "input[id*='captcha-answer']",
        "input[name*='quiz']",
        "input[type='text'][placeholder*='answer']",
        "input[type='number'][name*='captcha']"
    ];

    let question = null;
    for (const sel of QUESTION_SELECTORS) {
        const el = document.querySelector(sel);
        if (el && el.textContent && el.textContent.trim().length > 0) {
            question = el.textContent.trim();
            break;
        }
    }

    /* Fallback: scan the body for any small element whose text
       matches a math-captcha pattern. Tightly bounded so we don't
       false-positive on ad copy. */
    if (!question) {
        const candidates = document.querySelectorAll('label, p, span, div');
        const re = /(\d+|zero|one|two|three|four|five|six|seven|eight|nine|ten|eleven|twelve|thirteen|fourteen|fifteen|sixteen|seventeen|eighteen|nineteen|twenty)\s*(?:\+|-|\*|×|x|÷|\/|plus|minus|times|multiplied by|divided by)\s*(\d+|zero|one|two|three|four|five|six|seven|eight|nine|ten|eleven|twelve|thirteen|fourteen|fifteen|sixteen|seventeen|eighteen|nineteen|twenty)/i;
        for (const el of candidates) {
            if (el.children.length > 2) continue; /* skip wrappers */
            const text = (el.textContent || '').trim();
            if (text.length > 0 && text.length < 200 && re.test(text)) {
                question = text;
                break;
            }
        }
    }

    let inputSelector = null;
    for (const sel of ANSWER_SELECTORS) {
        if (document.querySelector(sel)) {
            inputSelector = sel;
            break;
        }
    }

    return { question, inputSelector };
})()
"#;

/// Solver for arithmetic math captchas.
pub struct MathCaptchaSolver;

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

impl MathCaptchaSolver {
    pub fn new() -> Self {
        Self
    }
}

#[derive(Debug, serde::Deserialize)]
struct ProbeResult {
    question: Option<String>,
    #[serde(rename = "inputSelector")]
    input_selector: Option<String>,
}

#[async_trait]
impl CaptchaSolver for MathCaptchaSolver {
    fn name(&self) -> &'static str {
        "MathCaptchaSolver"
    }

    fn method(&self) -> SolveMethod {
        SolveMethod::BehavioralBypass
    }

    fn supports(&self, kind: &DetectedCaptcha) -> bool {
        // Math captchas surface as: Custom("wp_math_captcha") (TOML
        // rule), PowCaptcha (some bundles route through it), or as a
        // generic Image/Canvas with embedded math text. Restrict
        // Custom to the known math-flavoured names so we don't
        // claim slider/click captchas like DataDome or PerimeterX.
        match kind {
            DetectedCaptcha::PowCaptcha
            | DetectedCaptcha::ImageCaptcha
            | DetectedCaptcha::CanvasCaptcha => true,
            DetectedCaptcha::Custom(name) => matches!(
                name.as_str(),
                "wp_math_captcha" | "math_captcha" | "math" | "arithmetic"
            ),
            _ => false,
        }
    }

    async fn solve(&self, page: &Page, _info: &CaptchaInfo) -> Result<CaptchaSolveResult> {
        let t0 = Instant::now();
        let raw = page
            .evaluate(PROBE_JS)
            .await
            .map_err(|e| anyhow!("math probe failed: {e}"))?;
        let probe: ProbeResult = raw.into_value().unwrap_or(ProbeResult {
            question: None,
            input_selector: None,
        });

        let (Some(question), Some(input_sel)) = (probe.question, probe.input_selector) else {
            return Ok(CaptchaSolveResult::failure(
                self.method(),
                t0.elapsed().as_millis() as u64,
            ));
        };

        let Some(answer) = solve_expression(&question) else {
            return Ok(CaptchaSolveResult::failure(
                self.method(),
                t0.elapsed().as_millis() as u64,
            ));
        };

        // Type the answer into the input. Use evaluate so we don't
        // depend on element handle lifetimes.
        let escape = serde_json::to_string(&answer.to_string()).unwrap_or_else(|_| "\"\"".into());
        let escape_sel =
            serde_json::to_string(&input_sel).unwrap_or_else(|_| "\"input[type=text]\"".into());
        let inject = format!(
            r#"
            (() => {{
                const el = document.querySelector({sel});
                if (!el) return false;
                el.value = {val};
                el.dispatchEvent(new Event('input', {{bubbles: true}}));
                el.dispatchEvent(new Event('change', {{bubbles: true}}));
                return true;
            }})()
            "#,
            sel = escape_sel,
            val = escape
        );
        let _ = page.evaluate(inject).await;

        let cookies = crate::cookies::capture_from_page(page)
            .await
            .unwrap_or_default();

        Ok(CaptchaSolveResult {
            solution: answer.to_string(),
            confidence: 1.0,
            method: self.method(),
            time_ms: t0.elapsed().as_millis() as u64,
            success: true,
            screenshot: None,
            cookies,
        })
    }
}

/// Parse + evaluate a single arithmetic expression embedded in
/// natural-language English. Returns `None` if no recognisable
/// `N OP N` shape is present.
pub(crate) fn solve_expression(text: &str) -> Option<i64> {
    // Normalise whitespace + lowercase. Keep the original case
    // around for word-number lookup which is case-insensitive
    // anyway.
    let t = text.to_lowercase();

    // Tokenise into number-words / digits / operators.
    // Replace word operators with symbols first.
    let t = t
        .replace("multiplied by", "*")
        .replace("divided by", "/")
        .replace(" plus ", " + ")
        .replace(" minus ", " - ")
        .replace(" times ", " * ")
        .replace("×", "*")
        .replace(" x ", " * ")
        .replace("÷", "/");

    let tokens: Vec<&str> = t
        .split(|c: char| !c.is_alphanumeric() && c != '+' && c != '-' && c != '*' && c != '/')
        .filter(|s| !s.is_empty())
        .collect();

    // Find first triple that looks like N OP N.
    for window in tokens.windows(3) {
        let a = parse_number(window[0]);
        let op = window[1];
        let b = parse_number(window[2]);
        if let (Some(a), Some(b)) = (a, b) {
            return apply(a, op, b);
        }
    }
    None
}

fn parse_number(s: &str) -> Option<i64> {
    if let Ok(n) = s.parse::<i64>() {
        return Some(n);
    }
    match s {
        "zero" => Some(0),
        "one" => Some(1),
        "two" => Some(2),
        "three" => Some(3),
        "four" => Some(4),
        "five" => Some(5),
        "six" => Some(6),
        "seven" => Some(7),
        "eight" => Some(8),
        "nine" => Some(9),
        "ten" => Some(10),
        "eleven" => Some(11),
        "twelve" => Some(12),
        "thirteen" => Some(13),
        "fourteen" => Some(14),
        "fifteen" => Some(15),
        "sixteen" => Some(16),
        "seventeen" => Some(17),
        "eighteen" => Some(18),
        "nineteen" => Some(19),
        "twenty" => Some(20),
        _ => None,
    }
}

fn apply(a: i64, op: &str, b: i64) -> Option<i64> {
    match op {
        "+" => Some(a + b),
        "-" => Some(a - b),
        "*" => Some(a * b),
        "/" if b != 0 => Some(a / b),
        _ => None,
    }
}

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

    #[test]
    fn solves_digit_addition() {
        assert_eq!(solve_expression("What is 3 + 5?"), Some(8));
    }

    #[test]
    fn solves_digit_subtraction() {
        assert_eq!(solve_expression("12 - 4 = ?"), Some(8));
    }

    #[test]
    fn solves_digit_multiplication() {
        assert_eq!(solve_expression("7 * 2 = ?"), Some(14));
        assert_eq!(solve_expression("7 × 2 = ?"), Some(14));
        assert_eq!(solve_expression("7 x 2 = ?"), Some(14));
    }

    #[test]
    fn solves_digit_division() {
        assert_eq!(solve_expression("20 / 4 = ?"), Some(5));
        assert_eq!(solve_expression("20 ÷ 4 = ?"), Some(5));
    }

    #[test]
    fn solves_word_numbers() {
        assert_eq!(solve_expression("What is two plus four?"), Some(6));
        assert_eq!(solve_expression("six minus three"), Some(3));
        assert_eq!(solve_expression("five times three"), Some(15));
    }

    #[test]
    fn solves_word_operators() {
        assert_eq!(solve_expression("4 plus 5"), Some(9));
        assert_eq!(solve_expression("10 minus 3"), Some(7));
        assert_eq!(solve_expression("3 times 4"), Some(12));
        assert_eq!(solve_expression("12 divided by 3"), Some(4));
        assert_eq!(solve_expression("5 multiplied by 6"), Some(30));
    }

    #[test]
    fn ignores_surrounding_text() {
        assert_eq!(
            solve_expression("Solve this anti-spam check: 8 + 7 = ? then submit"),
            Some(15)
        );
    }

    #[test]
    fn handles_no_match() {
        assert!(solve_expression("How are you today?").is_none());
        assert!(solve_expression("").is_none());
        assert!(solve_expression("just one number 5").is_none());
    }

    #[test]
    fn handles_div_by_zero() {
        assert!(solve_expression("5 / 0").is_none());
    }

    #[test]
    fn handles_negative_results() {
        assert_eq!(solve_expression("3 - 10"), Some(-7));
    }

    #[test]
    fn first_match_wins_when_multiple_expressions() {
        // First expression in source order is what we return.
        assert_eq!(
            solve_expression("Question 1: 2 + 2 = ? Question 2: 3 + 3 = ?"),
            Some(4)
        );
    }

    #[test]
    fn solves_zero_through_twenty_word_numbers() {
        for (i, word) in [
            "zero",
            "one",
            "two",
            "three",
            "four",
            "five",
            "six",
            "seven",
            "eight",
            "nine",
            "ten",
            "eleven",
            "twelve",
            "thirteen",
            "fourteen",
            "fifteen",
            "sixteen",
            "seventeen",
            "eighteen",
            "nineteen",
            "twenty",
        ]
        .iter()
        .enumerate()
        {
            assert_eq!(parse_number(word), Some(i as i64), "word: {word}");
        }
    }

    #[test]
    fn name_and_method_stable() {
        let s = MathCaptchaSolver::new();
        assert_eq!(s.name(), "MathCaptchaSolver");
        assert_eq!(s.method(), SolveMethod::BehavioralBypass);
    }

    #[test]
    fn supports_relevant_kinds_only() {
        let s = MathCaptchaSolver::new();
        assert!(s.supports(&DetectedCaptcha::Custom("wp_math_captcha".into())));
        assert!(s.supports(&DetectedCaptcha::Custom("math_captcha".into())));
        assert!(s.supports(&DetectedCaptcha::PowCaptcha));
        assert!(s.supports(&DetectedCaptcha::ImageCaptcha));
        assert!(s.supports(&DetectedCaptcha::CanvasCaptcha));
        // Slider/non-math Custom names must NOT match — math should
        // not claim DataDome or PerimeterX, which are interactive
        // slider/click captchas.
        assert!(!s.supports(&DetectedCaptcha::Custom("datadome".into())));
        assert!(!s.supports(&DetectedCaptcha::Custom("perimeterx_human".into())));
        assert!(!s.supports(&DetectedCaptcha::Custom("akamai_bot_manager".into())));
        assert!(!s.supports(&DetectedCaptcha::Turnstile));
        assert!(!s.supports(&DetectedCaptcha::None));
    }

    #[test]
    fn probe_js_includes_documented_selectors() {
        for sel in QUESTION_SELECTORS {
            assert!(
                PROBE_JS.contains(sel),
                "probe must contain question sel: {sel}"
            );
        }
        for sel in ANSWER_SELECTORS {
            assert!(
                PROBE_JS.contains(sel),
                "probe must contain input sel: {sel}"
            );
        }
    }
}