Bot Detection and Mitigation

11/7/2025

This post details how Nous Research implemented Vercel BotID Deep Analysis to protect their AI chat application during a free tier promotion. Key technical details include: layered checks at the portal (Auth Flow) before and after sign-up/sign-in, and at the chat UI layer with a heartbeat mechanism and cached verification for inference requests. It also highlights BotID's ability to detect and mitigate a coordinated attack characterized by specific IP patterns, geographic origin, and JA4 hashes, preventing a 3,000% traffic spike from impacting inference and availability.

10/31/2025

This post details the real-time detection and mitigation of a sophisticated bot network by Vercel BotID Deep Analysis. It highlights the system's ability to identify new browser profiles, correlate them across proxy nodes, and automatically re-verify sessions to confirm malicious activity. The process from detection to blocking took approximately 5 minutes without manual intervention. The post emphasizes the importance of adaptive, ML-based bot detection for handling advanced evasion tactics.

9/22/2025

This post details how Vercel's BotID, specifically its verified bot registry, was instrumental in diagnosing a case of SEO poisoning. By accurately identifying traffic as legitimate Google crawlers, the financial institution was able to avoid misinterpreting the situation and instead focus on submitting removal requests to Google Search Console for legacy indexed pages. The post highlights the importance of bot identification alongside blocking, emphasizing how understanding traffic sources can reveal deeper issues like technical debt.

8/13/2025

This post introduces and categorizes three types of AI bot traffic: AI training crawlers, AI engine grounding bots, and AI referrals. It explains the value of each type for discoverability and growth, and provides guidance on when and how to selectively block AI crawlers without harming discoverability. It also mentions Vercel's tools (Vercel Firewall, Bot Protection, BotID) as solutions for managing this traffic.

6/25/2025

Introduces BotID, a new layer of protection on Vercel for critical routes. BotID injects obfuscated JavaScript to collect signals distinguishing human users from bots, mutating detection logic on each load. It integrates with Kasada for Deep Analysis, providing enterprise-grade bot detection. The post details server-side verification with a single function call and highlights its application in protecting against scalpers, fake accounts, credential stuffing, and scraping.

5/23/2025

Introduced Bot Filter public beta, a managed WAF ruleset for advanced bot protection that can be activated with one click. This feature challenges non-browser bot traffic while allowing verified webhook providers and internal cron jobs. The post also details a significant increase in malicious traffic and IP addresses, with Vercel's WAF mitigating over 148 billion malicious requests from over 108 million unique IP addresses.

4/23/2025

Introduces the Bot Protection managed ruleset, a new security layer for the Vercel WAF that allows users to protect their applications from bot attacks with a single click. It uses heuristics to distinguish between human and bot traffic, allowing verified bots and challenging others. The post also mentions ongoing work on deeper observability into bot traffic, more granular control over verified and AI bots, a Bot Protection SDK, and runtime bot protection capabilities.

4/7/2025

Introduced Protectd, a next-generation real-time security engine that reduces DoS mitigation times by over tenfold. Protectd moves detection to the edge, replacing a polling-based design with real-time signal propagation. It uses Golang, a custom event-processing library, JIT and SIMD JSON serialization, virtualized time, and fast, probabilistic data structures to process ~550K events per second globally. New features include shadow mode for testing new rules, percentage-based rollouts, and historical traffic replay for backtesting. Since full rollout in January 2025, P99 time to mitigation is 3.5 seconds, P50 is 2.5 seconds, and response times can be as low as 0.5 seconds.

2/21/2025

This post details Vercel's security enhancements in Q4 2024, focusing on WAF and Firewall improvements. It highlights the mitigation of 87 billion attacks, a 40x speedup in DDoS protection via real-time stream processing, and the introduction of 403 Forbidden responses for persistent WAF threats. Visibility has been improved with DDoS mitigation notifications and more granular Firewall data in the Monitoring tab. Enterprise customers gain SIEM integration for audit logs, and Vercel has achieved PCI DSS v4.0 compliance.

1/24/2025

This post introduces Vercel's defensive playbook against Denial of Wallet (DoW) attacks, which aim to drain a service's operational budget by exploiting cloud scalability. It details how Vercel uses rate limiting, its Web Application Firewall (WAF) with challenge-response protocols, spend controls and real-time monitoring, and multifaceted bot protection (behavioral analysis, challenge-based verification, adaptive filtering) to mitigate these threats. The post emphasizes Vercel's secure-by-default approach and how these features protect both infrastructure and budget.

12/5/2024

This post details the layered security approach of Vercel, focusing on the Vercel Firewall, Deployment Protection, and the Web Application Firewall (WAF). It explains how the platform-wide Vercel Firewall provides automated, system-wide protection against DDoS attacks, malicious TCP connections, and other infrastructure abuses at network (Layer 3), transport (Layer 4), and application (Layer 7) layers. Deployment Protection offers project-level access control based on Vercel account, password protection, or trusted IPs. The WAF is described as a customizable layer with features like Attack Challenge Mode and a custom rules engine that allows granular traffic control using over 15 parameters (IP addresses, geolocation, user agents, headers) and actions (logging, challenges, blocking). It also highlights the integration of WAF persistent actions with the platform-wide firewall to stop repeat offenders without incurring usage costs.

10/2/2024

Introduced persistent actions for WAF rules to block repeat offenders for customizable durations, rate limiting for specific endpoints to prevent API abuse, and a Firewall API for programmatic control of WAF settings. Extended the Challenge page functionality to the broader Vercel Firewall. Highlighted instant rule propagation (300ms) and the integration of these features into the development workflow.

9/24/2024

This post details Vercel's multi-layered security approach, including Layer 4 and Layer 7 protections, application-aware routing, and real-time threat mitigation for infrastructure-level (bots, DDoS, SYN Floods) and application-level (SQL injection, API abuse) threats. It highlights the platform's ability to handle massive traffic surges, as demonstrated by mitigating a 1.37 Tbps DDoS attack with no downtime. It also introduces the ability to block AI bots using custom Firewall rules.

8/2/2024

This post details how Vercel's Firewall can be configured with custom rules to block or challenge malicious traffic based on IP address, JA4 fingerprints, and other criteria. It also introduces Attack Challenge Mode for temporary traffic verification and discusses real-time usage and spend controls, including spend limits and automatic app pausing. Code optimizations like enabling Fluid compute and transformation-based image optimization, and moving large media files to Vercel Blob storage are also highlighted as methods to protect applications and manage costs. The post also mentions the default enablement of deployment protection for new projects.

5/23/2024

Introduces the Vercel Web Application Firewall (WAF) as a first-party, Vercel-native security layer at the edge. Key features include a customizable rules engine (path, user agent, IP, geolocation, JA4 fingerprints), framework-aware rules, managed rulesets for Enterprise, observability into security metrics, beta rate limiting, instant propagation (300ms), and instant rollback. The WAF integrates with Vercel's existing DDoS mitigation and Attack Challenge Mode, leveraging framework-defined infrastructure for application-aware security decisions.

3/22/2024

This post details the integration of Kasada's bot protection technology with Vercel's AI SDK Playground. It highlights the threat vectors specific to AI applications, such as denial-of-wallet and prompt injection attacks. The implementation leverages Next.js Middleware to classify requests based on Kasada's bot classification, resulting in a significant reduction in bot traffic (from 84% to negligible levels). The post also touches upon Vercel's existing DDoS mitigation and Vercel Firewall capabilities as a first line of defense.

5/2/2023

Introduced Vercel Firewall with DDoS mitigation and custom firewall rules for IP blocking. Also introduced Vercel Secure Compute for Enterprise customers, providing private network connections for Serverless Functions and builds with dedicated IP addresses and logical separation.