[▲ Vercel Community](/) · [Categories](/categories) · [Latest](/latest) · [Top](/top) · [Live](/live) [Help](/c/help/9) # Pro plan: P1 bot traffic passing WAF challenges at scale 198 views · 0 likes · 3 posts Mikelabs (@mikelabs) · 2025-12-20 ## Current behavior * I’m seeing a **bot/scraper/DDOS-style spike** that ramps up fast. * In Vercel Firewall analytics, this traffic shows a **high “challenge solved” rate** and a large increase in **Allowed** requests. * If I set a rule to **Deny** the JA4 digest, it blocks **legit users** (it blocked me too), so I can’t leave it as Deny. * If I switch the JA4 rule to **Challenge**, it does **not** reduce the volume in a meaningful way. **Fingerprint involved:** * JA4 digest: `t13d1516h2_8daaf6152771_d8a2da3f94cd` * UA often seen (spoofable and tends to rotate): `Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/139.0.0.0 Safari/537.36` ## Expected behavior * A challenge should **materially reduce** automated request volume or fail more often under automation. * I need a mitigation that works even when: * IPs rotate (residential/proxy style) * paths are highly variable * I also need guidance that avoids blocking legit traffic. --- ## Code, configuration, and steps to reproduce ### Vercel Firewall configuration Plan: **Pro** No Cloudflare in front. Traffic hits Vercel directly. Managed protections: * AI Bots Managed Ruleset: **Enabled** Custom rules (in order): 1. **Rule: Match JA4 digest** * Condition: `JA4 Digest == t13d1516h2_8daaf6152771_d8a2da3f94cd` * Action tried: * `Deny` → blocks legitimate traffic (unacceptable) * `Challenge` → attacker still passes at high rate (like 20-50%, however attacker behavior is clearly automated) 2. **Attack Mode** * Enabled during active spike (temporary, does not work) * Result: traffic still shows high “challenge solved” and Allowed requests remain high ### Steps 1. Observe traffic spike in firewall / observability. 2. Create rule matching JA4 digest above. 3. Set action to **Challenge**. 4. (Optional) Enable **Attack Mode** during the spike. 5. Observe that: * challenge solve rate stays high * Allowed request volume stays high 6. Switch the rule to **Deny** and confirm it blocks legit users (so it’s not viable). Obviously you can't follow these if you're not attacked but I include them anyway --- ## Evidence / data (include what you can) Time window (ET): 12/18-12/19/25 Screenshots attached. Comical. Note that I enabled bot protection after I took this screenshot. It had no effect.  Support case reference: * Vercel support case: **889321** - My case has only gotten one canned responses from a support agent who didn't read it and didn't put any effort to respond in. --- ## Questions 1. Is Vercel seeing an abnormal “challenge solved” rate for this JA4 digest? 2. Is there a stronger mitigation when an actor can solve challenges at scale? 3. What rate limiting approach works best here (key choice + safe starting thresholds) given rotating IPs and variable paths? 4. Are there behavior/fingerprint controls beyond User-Agent that Vercel recommends? 5. Can Vercel apply any managed blocks server-side for this pattern? This is a really sophisticated and expensive attack because it chews up resources. Anshuman Bhardwaj (@anshumanb) · 2026-01-07 Hi there, I see that there's an ongoing support thread about this and our Support team experts are sharing solutions there. To avoid miscommunication or duplicates I'll defer to them. Mikelabs (@mikelabs) · 2026-01-13 They closed my ticket and did not resolve the issue, then issued me a refund. It would be nice if someone from the Vercel team could actually address this.
