Cloudflare Unveils 'Adaptive Intelligence' to Undermine the Economics of Bot Attacks

Cloudflare has unveiled a new bot detection engine called Adaptive Intelligence, stating its goal to resolve the economic imbalance between defense and attack. While existing defense systems relied on fixed rules that attackers could easily bypass, this engine focuses on continuously changing detection criteria to maximize the cost and time of attacks.

Limitations of Existing Bot Detection

Modern bot attacks have become organized and sophisticated. Attackers utilize residential proxy networks to distribute requests and disguise each request as a new visitor to bypass rate limits. Existing rule-based detection has a 'deterministic' characteristic where identical inputs always yield the same results, making it easy for attackers to identify system boundaries and learn bypass methods with automated tools. By the time defenders update their rules, attackers have already switched to new methods.

How Adaptive Intelligence Works

Instead of raising the 'wall' that blocks bots, Adaptive Intelligence adopts a strategy to make performing attacks uneconomical. To achieve this, it meets two core conditions.

• Non-deterministic detection: Instead of fixed rules, it statistically evaluates multiple signals. This prevents attackers from isolating specific logic to bypass detection.
• Feedback blocking: Even when bots are detected, the system does not react visibly, preventing attackers from knowing whether their bypass attempts failed or succeeded.

This approach renders information learned by attackers obsolete, designed so that attack costs increase as repeated attempts accumulate. Cloudflare has analyzed over 1 trillion requests per day and witnessed rapid changes in attack tactics, aiming to reverse the pace of adaptation on the defensive side to match the attackers through this engine.