Temporal - Bot or Human? Detecting Malicious Bots with Machine Learning in 2021

Explore the intricacies of detecting malicious bots using machine learning in this 23-minute OWASP Foundation conference talk. Delve into the challenges posed by sophisticated bot developers who design software to bypass detection systems, including their use of perfect browsers, mobile apps, and headless browsers. Learn about the complex techniques employed by bad bots, such as manipulating HTTP headers, changing browser fingerprints, and utilizing residential IPs. Discover the inner workings of a modern bot detection engine, including the collection and enrichment of server-side and client-side signals. Examine the challenges of authenticating good bots and detecting frameworks like Puppeteer extra stealth, Playwright, Selenium, and Headless Chrome. Gain insights into machine learning approaches for bad bot detection, with a focus on combining supervised and unsupervised techniques for maximum predictive accuracy. Understand key concepts such as automated threats, single-request attacks, residential proxies, and ongoing attack detection through real-world examples and case studies.

ABOUT DATADOME
AUTOMATED THREATS
AGENDA
BOTS & HACKERS TARGET THE WEAKEST LINK
PERFECT BROWSERS/APPS
REAL DEVICES
SINGLE-REQUEST ATTACKS
1/3 OF BAD BOTS USE RESIDENTIAL PROXIES
HOW DO THEY ACCESS CLEAN PROXIES?
HUMANS TO THE RESCUE
BOT SAAS SERVICES ARE NOT NEW
LUMINATI BECAME BRIGHT DATA
THE DATADOME BOT DETECTION ENGINE
DATADOME R&D REPORT
DETECT BROWSER AUTOMATION
STACKED MODEL PREDICTION
HOW TO APPLY ML FOR BOT DETECTION
SOLUTION
ONGOING ATTACK DETECTION
AVAILABLE DATA
INFERRING MALICIOUS FINGERPRINTS
BLOCKING PATTERN GENERATION (2)
SAFE BLOCKING PATTERNS
RESULTS: 28K MALICIOUS LOGIN ATTEM BLOCKED
KEY TAKEAWAYS