Description

Cybersecurity firm Wiz has identified five critical vulnerabilities in SAP AI Core, a cloud-based platform for AI workflows, collectively termed SAPwned. These flaws pose significant risks, allowing unauthorized access to customer data and internal artifacts, and potentially affecting cloud environments like AWS, Microsoft Azure, and SAP HANA Cloud. Attackers can exploit these vulnerabilities to modify Docker images and artifacts on SAP's internal registry, including the Google Container Registry and Artifactory server, threatening the software supply chain. The vulnerabilities arise from inadequate isolation and sandboxing in AI model training procedures. Exploiting misconfigurations in AWS Elastic File System (EFS) shares, attackers can access internal networks and sensitive data. They can gain cluster administrator privileges on SAP AI Core’s Kubernetes cluster by exploiting the Helm package manager's read/write access, enabling them to view and tamper with other customers' Pods, steal data, and manipulate AI models. This underscores the need for strict isolation and sandboxing to separate untrusted code from critical assets and for regular reviews of network configurations and access controls. Netskope reports a rise in enterprise generative AI use, highlighting the importance of robust data protection to prevent breaches. Concurrently, a new threat group, NullBulge, targets AI and gaming entities with malicious code and ransomware, underscoring the evolving cybersecurity landscape. This group exploits public repositories on platforms like GitHub, posing an ongoing threat to the software supply chain and emphasizing the need for proactive monitoring and response to potential cyber threats.