AI helps automatically fix cloud security problems. Company reports a 90 percent drop in critical risks

Tamnoon presented through an Amazon Web Services case study the results of Tami, a system that automates investigation and remediation of security problems in cloud infrastructure. Operators of large cloud environments use many tools to detect misconfigurations, dangerous permissions and other vulnerabilities. These systems can generate thousands of alerts that a security team must review one by one. Finding a problem, however, does not mean fixing it. An expert must determine which device or service is at risk, who is responsible for it and whether the proposed change could disrupt operations. Tamnoon therefore created a system composed of several specialized agents. They sequentially evaluate a security finding, add operational context, propose a solution and verify its safety. According to published figures, the system analyzed more than 14 million security alerts from customer environments. The company says that critical security risks can be reduced by as much as 90 percent within ninety days. Other reported results include up to a hundredfold improvement in efficiency when grouping alerts by common root causes and more than 13,000 working hours saved per week across the customer base. These figures come from Tamnoon’s research and operating data. They cannot automatically be treated as results that any organization can achieve. Technically, the system is built on Amazon Bedrock and other cloud services. Individual agents have separate tasks and their activity is controlled through secure interfaces. An important element is classification of proposed fixes. The system distinguishes safe interventions, risky changes and situations in which it lacks sufficient information. Depending on the result, it can continue automatically or hand the decision to a person. This approach is essential because an incorrect security fix can cause a service outage, loss of access or disruption of the production environment. Tamnoon says it has not observed production problems during automated remediation so far. This claim should be understood within the scope of the published case study, not as a guarantee of zero risk in future deployments. The system also uses access restrictions, separation of permissions and controls on content passed to models. Sensitive data are protected and individual services operate in isolated environments. The commercial significance lies in shifting security teams from manually triaging alerts toward addressing the actual causes of problems. Automation can reduce the time infrastructure remains exposed to a known risk. When deploying such a system, companies must track not only the number of alerts removed but also the severity of problems actually resolved, the number of incorrect interventions and the effect on service availability. The case shows that autonomous security systems can deliver measurable results when they have precisely defined authority, controlled tools and a way to hand risky decisions to a human. KEY TERMS Cloud security configuration: Settings for permissions, network access and other rules that protect services operated in the cloud. Vulnerability remediation: An action that removes or limits a specific security problem. Controlled autonomous intervention: An automatically executed change that must meet predefined safety conditions.

AI helps automatically fix cloud security problems. Company reports a 90 percent drop in critical risks

The full article is available to users registered at Hard Skills