Agentic AI Dramatically Accelerates Cloud Compromise to 72 Hours
Threat actors are leveraging agentic AI to reduce complex cloud compromise campaigns from weeks to mere days, according to a new Sygnia report.
A single threat actor was able to execute a sophisticated cloud attack in just 72 hours, a process that would typically take weeks, according to a new report by cybersecurity firm Sygnia. The report, titled "Inside an AI-Assisted Cloud Attack: Familiar Techniques at Unfamiliar Speed," highlights how attackers are increasingly using AI for speed and scale, rather than developing novel malware or zero-day exploits.
The compromised target was an Amazon Web Services (AWS) environment, with the attacker employing well-established techniques to gain access and achieve their objective of extortion. The actor exploited critical control gaps within the victim's secrets management, identity governance, deployment workflows, and cloud permissions. Initial access was gained through an internet-facing application that exposed an access key to one of the AWS accounts.
Once inside, the threat actor utilized AI-assisted or agentic workflows to concurrently perform four key tasks. These included a broad search for secrets and credentials across various layers of the AWS environment, such as plaintext secrets in S3 buckets, API keys in databases, and sensitive information stored in AWS Secrets Manager and Systems Manager Parameter Store. Simultaneously, the actor worked on establishing persistence mechanisms, including creating new access keys and IAM users, setting up reverse shells on EC2 instances and ECS containers, and altering deployment files.
Further concurrent actions involved exfiltrating data from RDS databases and executing "impact actions" to demonstrate their capabilities to the victim. These disruptive actions included denying access to S3 buckets, reducing ECS services to zero capacity, creating network-blocking ACL rules, and purging SQS queues. The report emphasizes that the organization's existing gaps in visibility, monitoring, identity controls, and incident preparedness significantly aided the attacker's rapid progress.
Avi Dayan, VP of incident response at Sygnia, noted that the most striking aspect of the attack was the speed and volume of malicious activity post-intrusion. "This case underscores a growing challenge for defenders: as large language models and agentic AI become more accessible, they have the potential to lower the barrier to entry, accelerate attack workflows, and enable less sophisticated or resource-constrained threat actors to operate with unprecedented speed and scale," Dayan stated.
Sygnia recommends several containment measures for network defenders to mitigate similar attacks. These include restricting cloud management access through IP allowlisting, disabling remote access VPNs until containment is complete, and restricting outbound internet connectivity for workloads to approved destinations only. Applying firewall policies and network access control lists (ACLs) to block known malicious infrastructure and restrict access to exposed assets is also crucial.
Additional recommended measures involve enforcing IP restrictions on source code repositories and development platforms, routing all application traffic through web application firewalls (WAFs), and implementing network segmentation and isolation controls to limit lateral movement. These proactive steps are vital in building a more resilient cloud security posture against increasingly sophisticated and accelerated threats.
The rapid advancement demonstrated in this attack signifies a new era in cyber warfare, where AI-powered tools can dramatically shorten the timeline from initial compromise to significant impact, potentially overwhelming traditional defense mechanisms and incident response capabilities.