Spandan Brahmbhatt

The Future of Scraper Protection in an AI-Agent-Driven Internet

Abstract: As AI agents increasingly automate web browsing, data comparison, and task-oriented workflows, non-human internet traffic is becoming significantly more sophisticated and pervasive. Traditional defense mechanisms—including IP reputation, rate limits, User-Agent validation, and basic headless browser checks—are rapidly losing standalone effectiveness against low-and-slow tactics, distributed residential proxies, and human-like automation frameworks. To address these evolving threats, cybersecurity systems must transition from binary human-versus-bot detection to evaluating whether an autonomous interaction can be trusted. This presentation introduces an "Agent Trust" framework centered around five core dimensions: Identity, Intent, Behavior, Consistency, and Impact. We outline a multi-layered technical detection stack incorporating network signals, device fingerprinting, behavioral telemetry, session-level analysis, population clustering, and predictive ML models. Additionally, the talk details an Agent Trust Scoring Model that categorizes incoming traffic—ranging from verified helpful agents and managed crawlers to spoofed agents and malicious scrapers—to trigger adaptive responses such as monitoring, rate limiting, challenging, deceiving, or blocking. Attendees will gain actionable strategies for defining modern automation policies, shifting from request-level detection to session-level traffic intelligence, and governing autonomous access in an agent-driven web. Profile: Lead Data Scientist | Cybersecurity & Behavioral Biometrics Expert Spandan Brahmbhatt is a Lead Data Scientist specializing at the intersection of cybersecurity, behavioral biometrics, and large-scale predictive modeling. Currently serving as Senior II Data Scientist at Akamai Technologies, he architects global detection systems that protect high-throughput digital ecosystems against advanced bot threats, unauthorized scraping, and fraud. With over 10 years of experience across leaders like Akamai, Arkose Labs, and Yum! Brands, Spandan holds a provisional patent in traffic anomaly forecasting and has led pioneering R&D in deep learning autoencoders and real-time AI agent identification. A former Northeastern University Lecturer with a Master’s in Information Systems, he is passionate about bridging academic AI research with production-grade MLOps and mentoring the next generation of data scientists.