If You're Not Using AI to Test Your Systems, Your Adversaries Will
Artificial intelligence plays a critical role in cybersecurity, both as a threat and a defense.
1The Role of AI in Cyber Attacks
Artificial intelligence has been effectively used in recent cyber attacks targeting organizations. AI creates a new attack surface for organizations trying to protect against both human and AI-based intrusions. Non-human identities that bypass traditional security policies further complicate this threat.
2Risks of Machine Identities
As AI transitions from generating content to taking actions, access to systems and sensitive data becomes inevitable. Matt Hartman, former head of the US Cybersecurity and Infrastructure Security Agency, emphasizes that every AI agent must be treated as a privileged identity, posing a significant management challenge for organizations.
3AI-Powered Identity Attacks
AI-powered identity and social engineering attacks are rapidly increasing. Techniques like highly personalized phishing and automated reconnaissance render traditional security measures ineffective. Hartman stresses the importance of focusing on strong authentication and zero-trust principles.
4Advantages of AI for Attackers
AI agents remain singularly focused on completing tasks without breaks. They excel at finding vulnerabilities, mapping networks, and identifying sensitive files. These capabilities provide a significant advantage for financially motivated criminals and state-backed cyber operatives.
5AI-Powered Red Team Testing
On the defense side, AI agents are used for red team testing. Former NSA cyber chief Rob Joyce warns that if organizations don't test their systems with AI, someone else will. AI-powered automated red team testing is becoming a critical need for every organization in the near future.
6Armadin's AI Attack Simulations
Kevin Mandia, founder of Mandiant, launched Armadin with $190 million in funding. The company develops thousands of AI agents to simulate real-life attackers. In a recent simulation, 17 million offensive actions were executed, and 38 attack paths were discovered.
7AI-Driven Security Findings
In Armadin's simulation, 101,169 alerts were analyzed with 100% accuracy across 231 billion raw events. This task would have taken a five-person analyst team approximately 2,400 hours to complete. AI significantly surpasses human capacity in such tasks.
8Expanding the Attack Surface with AI
Armadin co-founder Evan Peña highlights that AI agents expand the attack surface. AI enables organizations to cover more attack surfaces. Peña emphasizes that AI surpasses human capacity in security assessments.
9Safe AI Attack Testing
Peña stresses that organizations must conduct safe AI attacks against their own systems. OpenAI's attacks on Hugging Face demonstrate the importance of such tests. AI has become a critical tool for quickly identifying vulnerabilities.