March 2026 Summaries

David Huckaby discusses the importance of integrating runtime security with posture-based security to close gaps in cloud security. While posture-based security focuses on preventive measures by ensuring proper configurations and addressing known vulnerabilities, it is insufficient against dynamic and evolving threats like zero-day vulnerabilities, supply chain attacks, and insider threats. Runtime security complements this by providing real-time detection and mitigation of threats, acting as a last line of defense during active attacks, and offering full visibility into system activities. Huckaby emphasizes that a comprehensive cloud security strategy should involve both approaches, as they work together to provide a multi-layered defense, allowing organizations to adapt to new threats and maintain a secure environment. Sysdig, leveraging the open-source foundation of Falco, leads in offering advanced runtime protection, helping numerous Fortune 500 companies secure their cloud environments effectively.

In an era where cloud-native workloads are dynamic and constantly changing, traditional security measures are insufficient for protecting modern infrastructure. The text emphasizes the need for a cloud defense strategy that operates in real time, leveraging runtime insights to detect and respond to threats as they occur. Key components of this approach include achieving full-stack visibility to understand the context of activities across all layers of the technology stack, ensuring resilience and scalability of security measures to handle large volumes of data, and providing actionable context to enable swift and effective responses to security threats. By integrating runtime insights deeply into security platforms, organizations can enhance their ability to detect, investigate, and respond to threats in real time, thereby securing their cloud environments effectively.

Attacks on AI infrastructure are increasing, not through prompt manipulation as commonly assumed, but via the underlying systems supporting AI operations. This shift in attack vectors highlights the need for a dedicated focus on AI infrastructure security, as AI systems, though still evolving, manage massive and sensitive data, broadening the risk surface. AI's integration into modern enterprise involves diverse components such as generative and predictive systems, training and inference services, and cloud-based architectures, each with unique security challenges. The article underscores the importance of understanding the operational and architectural nuances of AI to effectively manage these risks, advocating for a comprehensive security approach that includes asset discovery, posture management, pipeline security, and runtime protection. The contrast between securing conversational inputs and the deeper infrastructure is likened to the difference between psychological and neurological assessments, emphasizing the need for robust infrastructure controls to prevent breaches.

Cloud security has evolved beyond mere visibility, as the rapid pace of AI-driven development and cloud-native applications necessitates a shift from posture-first security to a model focused on real-time action. While visibility was once sufficient to manage and mitigate risks in early cloud adoption, the current landscape requires security teams to understand not just configurations but also runtime contexts, linking activities such as logins, process executions, and network calls to form a coherent view of incidents. This approach enables more targeted responses, reducing the time and disruption associated with addressing security breaches. As attackers leverage AI to exploit vulnerabilities quickly, organizations must adopt runtime insights and AI-driven guidance to effectively manage risks and maintain control over their environments. This strategic shift is crucial for minimizing breach impacts, operational disruptions, and costs, as well as ensuring that organizations can keep pace with the evolving threat landscape.

AI coding agents are increasingly integrated into developer environments and CI/CD pipelines, operating with minimal oversight and posing unique security challenges due to their ability to execute commands, read files, and make network connections. These agents, which include Claude Code, Gemini CLI, and Codex CLI, are structurally vulnerable to prompt injections, operate with broad OS-level permissions, and lack robust separation between instruction and data, making them susceptible to manipulation. Current runtime security measures are inadequate as these agents behave more like interactive users rather than deterministic programs, necessitating syscall-level observation to detect unauthorized activities. The Sysdig Threat Research Team has developed detection strategies focusing on observable behaviors at the syscall level, such as unauthorized configuration access and safety control bypasses, to monitor these agents effectively. As these systems evolve and integrate further into cloud environments, the need for robust security measures becomes more critical to address their expanding attack surfaces and the complex threat models they introduce.

As AI coding agents like Claude Code, OpenAI’s Codex, and Gemini CLI become increasingly prevalent for accelerating development processes, they also introduce new security challenges that traditional tools are not equipped to handle. Sysdig addresses these concerns by offering runtime detections specifically designed for AI coding agents, providing real-time visibility into their behavior across developer and cloud environments. This capability enables security teams to differentiate between normal and potentially harmful activities, such as unauthorized file access or risky command executions. AI coding agents, powered by large language models, operate with extensive permissions, making them attractive targets for attacks that could result in remote code execution, credential theft, and supply chain attacks, among other risks. Sysdig's approach focuses on monitoring runtime actions rather than just the code itself, ensuring that organizations can securely adopt AI-assisted development while maintaining compliance and protecting sensitive data. By integrating these detections with existing security frameworks, Sysdig supports the secure scaling of AI technologies, emphasizing the importance of runtime security in an era where AI is deeply embedded in software development and operations.

TeamPCP, a threat actor, compromised Aqua Security's Trivy vulnerability scanner and Checkmarx's GitHub Actions to inject a credential-stealing payload into CI/CD pipelines, affecting numerous repositories. The attack involved force-pushing malicious commits to redirect actions, enabling credential scraping, cloud metadata harvesting, and encrypted exfiltration to typosquat domains. The Sysdig Threat Research Team identified a similar pattern in Checkmarx's GitHub Action shortly after the initial Trivy compromise, indicating a broader supply chain compromise. Runtime detection tools like Sysdig Secure and Falco proved effective in identifying the credential theft and exfiltration activities due to their focus on system behavior rather than relying on domain reputation or static analysis. This incident underscores the importance of runtime monitoring and using commit SHA pinning to prevent such attacks, as the same threat actor expanded their reach across the CI/CD ecosystem using deceptive techniques to avoid detection.

A critical vulnerability, CVE-2026-33017, in Langflow, an open-source AI framework, allowed unauthenticated remote code execution via a public endpoint, resulting in rapid exploitation within 20 hours of its disclosure. Attackers swiftly constructed working exploits based on the advisory, targeting exposed instances to exfiltrate sensitive information like keys and credentials, posing risks to connected databases and the software supply chain. Sysdig Threat Research Team observed the attack unfold in phases, from automated scanning to targeted data harvesting, using both automated tools and custom scripts. This incident highlights the urgent challenge for defenders as the time between vulnerability disclosure and exploitation has drastically shortened, emphasizing the need for runtime detection, network segmentation, and rapid response as traditional patch cycles prove insufficient against such swift threats.

Sysdig's Inline Cloud Response for AWS enhances the efficiency of Security Operations Center (SOC) and Computer Security Incident Response Team (CSIRT) workflows by integrating AWS-native investigative and containment actions directly into the Sysdig console. This approach reduces the time and complexity typically involved in threat detection and response, allowing analysts to act swiftly without needing to switch tools or coordinate across multiple teams. By streamlining the process of identifying, investigating, and containing threats, Sysdig minimizes operational delays and miscommunications, ultimately decreasing the mean time to containment (MTTC) and reducing the potential business impact of security incidents. This system empowers security teams to address threats in real-time, maintaining focus on threat mitigation rather than being bogged down by procedural overhead, thus transforming the traditional observe-and-escalate model into a proactive, inline response strategy.

On March 9, 2026, a critical fix for CVE-2026-3288 was merged into the Kubernetes ingress-nginx project to address a configuration injection vulnerability in the NGINX Ingress Controller, allowing any user with permission to create or modify Ingress resources to inject arbitrary nginx configuration directives by manipulating the Ingress path field. This vulnerability can lead to remote code execution and the disclosure of secrets, posing significant security risks. It is closely related to CVE-2026-24512, which was addressed earlier but left certain areas vulnerable due to incomplete sanitization efforts, specifically in the buildProxyPass() function. The Sysdig Threat Research Team analyzed these vulnerabilities and developed a Falco detection rule to identify exploitation attempts through Kubernetes audit logs. Organizations are advised to update to the latest fixed versions, v1.13.8, v1.14.4, or v1.15.0, and implement detection mechanisms to mitigate the risks. This issue underscores the ongoing challenges of safely handling user-controlled input within nginx configuration templates, as seen in previous ingress-nginx vulnerabilities.

Runtime malware detection for AWS Fargate has been enhanced by Sysdig to accommodate serverless container environments, allowing security teams to maintain consistent threat detection and response capabilities across modern cloud infrastructures. As organizations adopt AWS Fargate for its operational efficiency in deploying containerized applications without managing the underlying infrastructure, the challenge becomes ensuring runtime security visibility remains intact. Traditional container security relies on host-level instrumentation, which is not feasible with serverless models like AWS Fargate. Sysdig's extension of its malware detection capabilities into these environments addresses the need for robust runtime threat detection, enabling security operations teams to investigate and respond to threats effectively. This approach supports platform modernization efforts by ensuring security visibility and compliance are not compromised, even as infrastructure management shifts away from traditional models. As cloud architectures evolve, maintaining runtime security visibility and detection coverage becomes crucial for organizations adopting serverless solutions, ensuring that they can detect and neutralize threats in real-time across diverse computing platforms.

Sysdig's cloud detection and response system offers a comprehensive solution for malware detection and prevention in modern cloud environments, addressing the limitations of traditional endpoint-based models. As malware in the cloud has become faster and more automated, Sysdig employs a layered approach combining execution-level detection, runtime blocking, and YARA-based pattern matching to protect cloud-native, hybrid, and on-premises environments. This approach allows for the identification and blocking of malicious binaries as they are written or executed, thus preventing potential threats from escalating. Sysdig's solution is optimized for containerized and dynamic workloads, reducing the operational burden associated with legacy scanning methods and ensuring consistent protection across diverse infrastructures. The Sysdig Threat Research Team continuously updates detection rules and intelligence feeds, enhancing the system's ability to counter both known and emerging threats efficiently. By transforming malware detection into an integrated, proactive defense capability, Sysdig enables security teams to make faster decisions, reduce operational risks, and maintain strong runtime protection across various architectures.

In February 2026, cybersecurity was dominated by issues related to artificial intelligence (AI), highlighting both its potential and the security risks it poses. AI-driven attacks, such as AI recommendation poisoning and AI-assisted ransomware, underscored the vulnerabilities in systems due to weak credentials and exposed management ports. Significant incidents included the rapid exploitation of the BeyondTrust vulnerability (CVE-2026-1731) and the compromise of Fortinet FortiGate devices across 55 countries. The month also saw AI being used to compress attack kill chains, as demonstrated by the Sysdig Threat Research Team, which showed how attackers could gain admin privileges in just eight minutes. Meanwhile, the integrity of AI systems' training data was called into question due to hidden manipulations by legitimate organizations. Despite the rapid evolution of AI threats, the importance of traditional security measures such as credential management, patching, and identity hygiene remains crucial, as emphasized by the need for real-time detection over reactive measures. The European Commission and the French Ministry of Finance dealt with separate cyber incidents, while companies like Anthropic and Sysdig introduced AI tools for security enhancement, even as the misuse of AI continues to grow.