stirfry-spam-filter/relay-spam-filter.js

#!/usr/bin/env node

const fs = require('fs');

// Configuration
const config = {
    fastReplyThreshold: 3, // 3 seconds
    maxRepliesPerMinute: 50,
    maxRepliesPerHour: 400,
    recentEventWindow: 3600, // 1 hour in seconds
    contentSimilarityThreshold: 0.8,
    relayUrl: "freelay.sovbit.host",
    
    // Periodic cleanup intervals
    cleanupInterval: 300, // 5 minutes
    statsRetentionPeriod: 86400, // 24 hours
    
    // Enhanced reputation system
    reputationConfig: {
        initialScore: 100,
        goodEventBonus: 1,
        spamPenalty: -15,         // Base penalty
        recoveryRate: 3,          // Base recovery rate
        minScore: -100,           // Absolute minimum
        maxScore: 1000,
        trustThreshold: 50,
        blockThreshold: -50,      // Start blocking at this point
        blockRecoveryThreshold: -25,  // Must recover to this to post again
        
        // Progressive penalties
        penaltyScaling: {
            0: -15,    // Normal penalty above 0 reputation
            "-25": -20, // Harsher penalty below -25
            "-50": -25  // Even harsher below -50
        },
        
        // Recovery scaling
        recoveryScaling: {
            0: 3,      // Normal recovery above 0
            "-25": 2,  // Slower recovery below -25
            "-50": 1   // Very slow recovery below -50
        }
    },

    // Kind-specific limits
    kindSpecificLimits: {
        0: { // Profile updates
            maxPerHour: 10
        },
        3: { // Contacts
            maxPerHour: 5
        },
        1: { // Regular posts
            maxPerMinute: 30,
            maxPerHour: 200
        }
    },

    // New user limits
    newPubkeyReplyThreshold: 60,
    newPubkeyMaxPostsIn5Min: 10,

    allowedKinds: [3, 5, 10001, 10002, 30311, 10050],
    bypassAllChecksKinds: [38383],
};

// In-memory stores with cleanup
const recentEvents = [];
const authorStats = new Map();
const authorReputation = new Map();
const newPubkeyPostCount = new Map();
const userViolationHistory = new Map();  // Track violation history

function log(message) {
    console.error(`[${new Date().toISOString()}] ${message}`);
}

function recordViolation(pubkey, eventKind, reason, penalty) {
    if (!userViolationHistory.has(pubkey)) {
        userViolationHistory.set(pubkey, []);
    }
    
    const history = userViolationHistory.get(pubkey);
    history.push({
        timestamp: Date.now(),
        eventKind: eventKind,
        reason: reason,
        penalty: penalty,
        currentReputation: getAuthorReputation(pubkey).score
    });

    // Keep only last 50 violations
    if (history.length > 50) {
        history.shift();
    }
}

function getAuthorReputation(pubkey) {
    if (!authorReputation.has(pubkey)) {
        authorReputation.set(pubkey, {
            score: config.reputationConfig.initialScore,
            lastUpdate: Date.now()
        });
    }
    return authorReputation.get(pubkey);
}

function updateReputation(pubkey, isSpam, eventKind, spamReason) {
    const reputation = getAuthorReputation(pubkey);
    const hoursSinceLastUpdate = (Date.now() - reputation.lastUpdate) / (1000 * 60 * 60);
    
    // Get appropriate recovery rate based on current score
    let recoveryRate = config.reputationConfig.recoveryRate;
    for (const [threshold, rate] of Object.entries(config.reputationConfig.recoveryScaling)
        .sort((a, b) => Number(b[0]) - Number(a[0]))) {
        if (reputation.score <= Number(threshold)) {
            recoveryRate = rate;
            break;
        }
    }
    
    // Apply scaled recovery
    reputation.score = Math.min(
        config.reputationConfig.maxScore,
        reputation.score + (hoursSinceLastUpdate * recoveryRate)
    );
    
    if (isSpam) {
        // Get appropriate penalty based on current score
        let penalty = config.reputationConfig.spamPenalty;
        for (const [threshold, pen] of Object.entries(config.reputationConfig.penaltyScaling)
            .sort((a, b) => Number(b[0]) - Number(a[0]))) {
            if (reputation.score <= Number(threshold)) {
                penalty = pen;
                break;
            }
        }
        reputation.score += penalty;
        
        // Record the violation
        recordViolation(pubkey, eventKind, spamReason, penalty);
    } else {
        reputation.score += config.reputationConfig.goodEventBonus;
    }
    
    reputation.score = Math.max(
        config.reputationConfig.minScore,
        Math.min(config.reputationConfig.maxScore, reputation.score)
    );
    
    reputation.lastUpdate = Date.now();
    return reputation.score;
}

function canUserPost(pubkey, eventKind) {
    const reputation = getAuthorReputation(pubkey);
    
    if (reputation.score <= config.reputationConfig.blockThreshold) {
        if (reputation.score < config.reputationConfig.blockRecoveryThreshold) {
            // Get violation history
            const history = userViolationHistory.get(pubkey) || [];
            const recentViolations = history
                .slice(-5)  // Get last 5 violations
                .map(v => `\n    - ${new Date(v.timestamp).toISOString()}: kind ${v.eventKind}, ${v.reason} (penalty: ${v.penalty})`)
                .join('');

            log(`User ${pubkey} blocked from posting: reputation (${reputation.score.toFixed(2)}) below recovery threshold
Recent violations:${recentViolations || '\n    No recent violations recorded'}`);
            return false;
        }
    }
    return true;
}

function getKindSpecificLimits(kind) {
    return config.kindSpecificLimits[kind] || {
        maxPerMinute: config.maxRepliesPerMinute,
        maxPerHour: config.maxRepliesPerHour
    };
}

function updateAuthorStats(pubkey, timestamp, kind) {
    if (config.bypassAllChecksKinds.includes(kind)) return;

    const now = Date.now();
    if (!authorStats.has(pubkey)) {
        authorStats.set(pubkey, {
            lastReply: now,
            repliesLastMinute: 0,
            repliesLastHour: 0,
            lastReset: now,
            kinds: new Map()
        });
    }

    const stats = authorStats.get(pubkey);
    const timeSinceLastReset = now - stats.lastReset;
    
    // Reset counters if needed
    if (timeSinceLastReset >= 3600000) { // 1 hour
        stats.repliesLastHour = 0;
        stats.repliesLastMinute = 0;
        stats.lastReset = now;
        stats.kinds.clear();
    } else if (timeSinceLastReset >= 60000) { // 1 minute
        stats.repliesLastMinute = 0;
    }

    // Update kind-specific counters
    if (!stats.kinds.has(kind)) {
        stats.kinds.set(kind, { hourly: 0, lastReset: now });
    }
    const kindStats = stats.kinds.get(kind);
    if (now - kindStats.lastReset >= 3600000) {
        kindStats.hourly = 0;
        kindStats.lastReset = now;
    }
    kindStats.hourly++;

    stats.repliesLastMinute++;
    stats.repliesLastHour++;
    stats.lastReply = now;
}

function isRateLimitExceeded(pubkey, kind) {
    const stats = authorStats.get(pubkey);
    if (!stats) return false;

    const kindLimits = getKindSpecificLimits(kind);
    const kindStats = stats.kinds.get(kind);
    
    // Consider reputation in rate limiting
    const reputation = getAuthorReputation(pubkey);
    const reputationMultiplier = Math.max(0.5, Math.min(2, reputation.score / 100));
    
    const adjustedMinuteLimit = kindLimits.maxPerMinute * reputationMultiplier;
    const adjustedHourLimit = kindLimits.maxPerHour * reputationMultiplier;

    const exceeded = 
        stats.repliesLastMinute > adjustedMinuteLimit ||
        stats.repliesLastHour > adjustedHourLimit ||
        (kindStats && kindStats.hourly > kindLimits.maxPerHour);

    if (exceeded) {
        log(`Rate limit exceeded for ${pubkey} (reputation: ${reputation.score.toFixed(2)}, kind: ${kind})`);
    }

    return exceeded;
}

function isFastReply(event, recentEvents) {
    const replyTo = event.tags.find(tag => tag[0] === 'e');
    if (replyTo) {
        const originalEvent = recentEvents.find(e => e.id === replyTo[1]);
        if (originalEvent && event.created_at - originalEvent.created_at <= config.fastReplyThreshold) {
            log(`Fast reply detected: ${event.created_at - originalEvent.created_at} seconds to ${originalEvent.id}`);
            return true;
        }
    }
    return false;
}

function calculateSimilarity(str1, str2) {
    const len1 = str1.length;
    const len2 = str2.length;
    const maxLength = Math.max(len1, len2);
    if (maxLength === 0) return 1.0;
    return (maxLength - levenshteinDistance(str1, str2)) / maxLength;
}

function levenshteinDistance(str1, str2) {
    const m = str1.length;
    const n = str2.length;
    const dp = Array.from({ length: m + 1 }, () => Array(n + 1).fill(0));

    for (let i = 0; i <= m; i++) dp[i][0] = i;
    for (let j = 0; j <= n; j++) dp[0][j] = j;

    for (let i = 1; i <= m; i++) {
        for (let j = 1; j <= n; j++) {
            if (str1[i - 1] === str2[j - 1]) {
                dp[i][j] = dp[i - 1][j - 1];
            } else {
                dp[i][j] = Math.min(
                    dp[i - 1][j - 1],
                    dp[i - 1][j],
                    dp[i][j - 1]
                ) + 1;
            }
        }
    }

    return dp[m][n];
}

function isContentCopy(event, recentEvents) {
    if (config.allowedKinds.includes(event.kind)) {
        log(`Skipping content copy check for allowed kind ${event.kind}`);
        return false;
    }

    for (const recentEvent of recentEvents) {
        if (recentEvent.pubkey !== event.pubkey) {
            const similarity = calculateSimilarity(recentEvent.content, event.content);
            if (similarity >= config.contentSimilarityThreshold) {
                log(`Content copy detected: ${similarity.toFixed(2)} similarity between ${event.id} and ${recentEvent.id}`);
                return true;
            }
        }
    }
    return false;
}

function isBotBehavior(event) {
    const isBotLike = event.content.endsWith(config.relayUrl);
    if (isBotLike) {
        log(`Bot behavior detected in event ${event.id}: content ends with relay URL`);
    }
    return isBotLike;
}

function isNewPubkeySpam(event) {
    const isNewPubkey = !authorStats.has(event.pubkey);
    if (isNewPubkey) {
        const replyTo = event.tags.find(tag => tag[0] === 'e');
        if (replyTo) {
            const originalEvent = recentEvents.find(e => e.id === replyTo[1]);
            if (originalEvent && event.created_at - originalEvent.created_at <= config.newPubkeyReplyThreshold) {
                log(`New pubkey ${event.pubkey} replied too quickly to ${originalEvent.id}`);
                return true;
            }
        }
    }
    return false;
}

function isRapidNewPubkeyPosting(event) {
    if (authorStats.has(event.pubkey)) {
        return false;
    }

    const now = Math.floor(Date.now() / 1000);
    const count = newPubkeyPostCount.get(event.pubkey) || 0;
    newPubkeyPostCount.set(event.pubkey, count + 1);
    
    if (count >= config.newPubkeyMaxPostsIn5Min && now - event.created_at <= 300) {
        log(`Rapid new pubkey posting detected for ${event.pubkey}: ${count} posts in 5 minutes`);
        return true;
    }
    
    if (now - event.created_at > 300) {
        newPubkeyPostCount.delete(event.pubkey);
    }
    
    return false;
}

function isSpam(event, recentEvents) {
    const pubkey = event.pubkey;
    
    log(`Checking event ${event.id} from pubkey ${pubkey} (kind: ${event.kind})`);

    if (config.bypassAllChecksKinds.includes(event.kind)) {
        log(`Bypassing all checks for kind ${event.kind}`);
        return { isSpam: false };
    }

    if (isRateLimitExceeded(pubkey, event.kind)) {
        return { isSpam: true, reason: `Rate limit exceeded for kind ${event.kind}` };
    }

    if (isNewPubkeySpam(event)) {
        return { isSpam: true, reason: 'New pubkey replying too quickly' };
    }

    if (isRapidNewPubkeyPosting(event)) {
        return { isSpam: true, reason: 'Rapid-fire posting from new pubkey' };
    }

    if (isFastReply(event, recentEvents)) {
        return { isSpam: true, reason: 'Fast reply to recent event' };
    }
    
    if (isContentCopy(event, recentEvents)) {
        return { isSpam: true, reason: 'Content copy detected' };
    }
    
    if (isBotBehavior(event)) {
        return { isSpam: true, reason: 'Bot-like behavior detected' };
    }

    log(`Event ${event.id} passed all spam checks`);
    return { isSpam: false };
}

function periodicCleanup() {
    const now = Date.now();
    
    // Clean up old events
    while (recentEvents.length > 0 && 
           (now - recentEvents[recentEvents.length - 1].created_at * 1000) > config.recentEventWindow * 1000) {
        recentEvents.pop();
    }
    
    // Clean up old author stats
    for (const [pubkey, stats] of authorStats.entries()) {
        if (now - stats.lastReply > config.statsRetentionPeriod * 1000) {
            authorStats.delete(pubkey);
        }
    }
    
    // Clean up old reputation data
    for (const [pubkey, rep] of authorReputation.entries()) {
        if (now - rep.lastUpdate > config.statsRetentionPeriod * 1000) {
            authorReputation.delete(pubkey);
        }
    }
    
    // Clean up newPubkeyPostCount
    for (const [pubkey, timestamp] of newPubkeyPostCount.entries()) {
        if (now - timestamp > 300000) { // 5 minutes
            newPubkeyPostCount.delete(pubkey);
        }
    }

    // Clean up old violation history
    for (const [pubkey, history] of userViolationHistory.entries()) {
        if (history.length === 0 || now - history[history.length - 1].timestamp > config.statsRetentionPeriod * 1000) {
            userViolationHistory.delete(pubkey);
        }
    }
}

const rl = require('readline').createInterface({
    input: process.stdin,
    output: process.stdout,
    terminal: false
});

// Set up periodic cleanup
setInterval(periodicCleanup, config.cleanupInterval * 1000);

rl.on('line', (line) => {
    let req;
    try {
        req = JSON.parse(line);
    } catch (error) {
        log(`Error parsing JSON: ${error.message}`);
        return;
    }

    if (req.type !== 'new') {
        log("unexpected request type");
        return;
    }

    let res = { id: req.event.id };

    // Check if user is allowed to post before any other processing
    if (!canUserPost(req.event.pubkey, req.event.kind)) {
        res.action = 'reject';
        res.msg = 'rejected: reputation too low to post';
        log(`Rejected event ${req.event.id} (kind: ${req.event.kind}): reputation too low`);
        console.log(JSON.stringify(res));
        return;
    }

    // Update recent events
    recentEvents.push(req.event);
    recentEvents.sort((a, b) => b.created_at - a.created_at);
    
    // Process event
    updateAuthorStats(req.event.pubkey, req.event.created_at, req.event.kind);
    
    const spamCheck = isSpam(req.event, recentEvents);
    const reputationScore = updateReputation(
        req.event.pubkey, 
        spamCheck.isSpam, 
        req.event.kind, 
        spamCheck.reason
    );

    if (spamCheck.isSpam) {
        res.action = 'reject';
        res.msg = `rejected: ${spamCheck.reason}`;
        log(`Rejected event ${req.event.id} (kind: ${req.event.kind}): ${res.msg} (reputation: ${reputationScore.toFixed(2)})`);
    } else {
        res.action = 'accept';
        log(`Accepted event ${req.event.id} (kind: ${req.event.kind}, reputation: ${reputationScore.toFixed(2)})`);
    }

    console.log(JSON.stringify(res));
});

log("Strfry filter started with config: " + JSON.stringify(config, null, 2));