Cloud ForeverDays - Cloud Vulnerabilities That Refuse To Die

The talk argues that cloud platforms keep carrying forward security weaknesses that were never really fixed, while adding opaque dependencies, concentrated infrastructure and AI-generated code. The speaker opens with how nation-state operators get malware onto devices, pointing to advertising-based zero-click delivery and to locked-down mobile and cloud platforms that limit what testers can examine.

The middle section covers supply-chain risk, from vendor backdoors and the XZ Utils incident to the danger of many services sharing the same underlying software and providers. He describes Office 365 password spraying that generates no logs, and argues that Active Directory flaws have been inherited by Entra ID. He also shows how blocklisted IP data grouped by ASN can reveal risky cloud neighborhoods, and walks through post-exploitation with GraphRunner, along with AzureHound data imported into BloodHound to expose paths to administrator rights.

The closing portion addresses AI-assisted development, which he links to a flattening of code quality and to business-logic flaws, and the frustration of reporting issues to large vendors. He recommends good-faith security research as a framework for working with providers, encourages practitioners to run these tools in their own environments, and points to lower-cost training as a way to widen access to the field.