Garbage Responder

​

Overview

​

Configuration

​

HTTP Response

​

Example Response Body

florb~
wibble^
5kJ3@xP9#mL2$nH7%fG1&dB4*aC8(eR6)iT0+
snark|
quint}
F8@aS3#jK9$xM2%nL7&pQ1*vB6(cD4)eR5+yT0-zW8/gH3<uI9>wN2?
blarg\
ploosh!
3L#9K@5J$2M&7N%1P*6Q(4R)8S+0T-5U/9V<2W>7X?1Y\3Z|4A}6B~8C
zaxor`
...

​

Examples

localhost:8080 {
    defender garbage {
        ranges openai aws deepseek
    }
    respond "Legitimate content for humans"
}

​

Implementation Details

func (g GarbageResponder) ServeHTTP(w http.ResponseWriter, _ *http.Request, _ caddyhttp.Handler) error {
    garbage := generateTerribleText(100)
    w.Header().Set("Content-Type", "text/plain")
    w.WriteHeader(http.StatusOK)
    _, err := w.Write([]byte(garbage))
    return err
}

​

Garbage Generation Algorithm

1. Nonsense Words - Random selection from a predefined list:
   • florb, zaxor, quint, blarg, wibble, fizzle, gronk, snark, ploosh, dribble
2. Random Characters - Random length (10-60 chars) from character set:
   • Lowercase letters: a-z
   • Uppercase letters: A-Z
   • Numbers: 0-9
   • Symbols: !@#$%^&*()_+-=[]{};':",./<>?\|~

​

Use Cases

​

AI Training Poisoning

defender garbage {
    ranges openai deepseek mistral
}

​

Scraper Resource Waste

defender garbage {
    ranges scrapers bots
}

​

Content Protection

defender garbage {
    ranges aws gcloud azure
}

​

Advantages

1. AI Poisoning - Degrades quality of training data if scraped
2. Unpredictable - Each response is randomly generated
3. Resource Waste - Scrapers waste bandwidth, storage, and processing
4. Looks Valid - Returns 200 OK so scrapers think they succeeded
5. No Blocking Signal - Scrapers don’t know they’re being served garbage

​

Disadvantages

1. Bandwidth Cost - Sends ~5-10KB per request instead of small error
2. Processing Overhead - Generates random data for each request
3. Still Allows Access - Doesn’t actually prevent scraping, just poisons it
4. May Be Detected - Sophisticated scrapers might detect random patterns

​

Comparison with Other Responders

• vs Block: Garbage returns 200 OK with data, Block returns 403 error
• vs Drop: Garbage sends a response, Drop terminates connection
• vs Custom: Garbage returns random data, Custom returns your message
• vs Tarpit: Garbage sends data quickly, Tarpit sends slowly

​

When to Use Garbage

• You want to poison AI training datasets
• Wasting scraper resources is a goal
• You want scrapers to think they succeeded
• Content protection is more important than bandwidth

• Bandwidth costs are a concern
• You want to explicitly block access
• You need to conserve server resources
• Clear error messages are desired

​

Best Practices

1. Combine with serve_ignore - Also serve robots.txt to discourage polite bots
2. Target AI services specifically - Use ranges like openai, deepseek, mistral
3. Monitor bandwidth - Garbage uses more bandwidth than error responses
4. Consider Tarpit - For even more resource waste, use Tarpit instead
5. Test with whitelist - Ensure legitimate users aren’t getting garbage

​

Testing

# From a blocked IP, you should see garbage
curl http://example.com

# Each request should return different garbage
curl http://example.com
curl http://example.com

# Simulate blocked IP using X-Forwarded-For
curl -H "X-Forwarded-For: 20.202.43.67" http://example.com

​

Ethical Considerations

• Protect your content from unauthorized scraping
• Degrade the quality of AI models trained on scraped data
• Waste resources of unauthorized scrapers

​

Related Documentation

• Tarpit Responder - Slow garbage delivery for maximum resource waste
• Block Responder - Explicitly deny access instead
• Custom Responder - Return a clear message instead
• serve_ignore Configuration - robots.txt directive