Cloudflare tests its WAF against advanced AI techniques
Cloudflare evaluated its web application firewall (WAF) by having large language models imitate hackers, generating and adapting attack payloads in real time. In over 1,100 simulated attacks across six categories, most attempts were blocked, but some variations bypassed the WAF. Cloudflare used these findings to develop new detection rules, improving its security for all customers.
- Cloudflare used LLMs to simulate adaptive cyberattacks
- Over 1,100 attack attempts were run against its WAF
- Most attacks were blocked, but some bypassed defenses
- Findings led to improved security rules for all customers
Sources covering this
More in Cybersecurity
Critical zero-days in Citrix NetScaler exploited, patches released
Citrix has confirmed that two critical zero-day vulnerabilities affecting NetScaler ADC and Gateway devices have been exploited.
Dutch police arrest former hacker in ShinyHunters probe
Dutch police have arrested Pepijn van der Stap, a previously convicted hacker, in connection with an ongoing investigation into the…
Apple fixes CoreGraphics flaw after targeted attacks
Apple has released security updates for iOS, iPadOS, and macOS to fix a CoreGraphics vulnerability that may have been used in highly…
New Spectre BTR attack exposes Linux root password hashes
Researchers have unveiled a new Spectre Variant 2 attack, named Branch Target Reuse (BTR), that can extract root password hashes from…