Parse notes
Sponsored
Modern applications and APIs are fueling rapid digital transformation—but they're also expanding attack surfaces. As threats evolve, enterprises must rethink their web application and API
This Gartner report [4] provides critical guidance for navigating today's WAAP landscape, including:
The Port of Seattle experienced a ransomware attack in August 2024 that disrupted its websites and phone systems, including Seattle-Tacoma International Airport. The breach affected the personal data of 90,000 individuals, prompting the Port to offer free credit monitoring and notify those impacted. The responsible group, Rhysida, has previously targeted multiple industries and organizations.
Oracle is warning customers that a hacker broke into a computing system twice in one month. Customer data, including login data, was stolen from a legacy environment that has not been used for eight years. However, customer log-in credentials from as recently as 2024 were stolen.
Researchers from Socket have discovered a malicious PyPi package called disgrasya. The tool abuses legitimate WooCommerce sites to validate cards using an attacker-controlled site masquerading as CyberSource. This package is particularly notable in that it does not abuse typosquatting or any other evasion technique but is instead overtly malicious.
Google launched a stable version of its model-signing library with NVIDIA and HiddenLayer to secure the machine learning supply chain. The library enables users to verify that a model used in an application is the one created by developers using digital signatures like those from Sigstore. This counters security threats from the multi-stage ML development (training, fine-tuning, and deployment), where malicious actors can tamper with models via weight alterations, backdoors, or code injection.
PowerUserAccess is a broad managed permission for AWS that explicitly restricts a user's access to some actions related to IAM, Organizations, and Account. Attackers can often accomplish one-time goals such as crypto mining or ransomware with PowerUserAccess but would require AdministratorAccess to establish persistence. However, in practice, attackers can often find an overly permissive resource to exploit to escalate privileges due to PowerUserAccess violating the principle of least privilege.
Signature verification and image signing enhance supply chain resiliency by ensuring image integrity. DataDog created a signing service for signing images to facilitate adoption across varied CI/CD processes that stored signatures in an OCI-compliant registry and verified these signatures. It managed signature verification during image pulls as a containerd plugin rather than through an admission webhook to prevent cluster-level dependencies in the Kubernetes control plane's hot path.
Google has launched Sec-Gemini v1, an experimental AI model that boosts cybersecurity. It merges advanced features with real-time knowledge, outperforming other models in key benchmarks. Sec-Gemini v1 empowers defenders with detailed threat analysis and vulnerability insights.
Enola is a Golang CLI tool that helps you find accounts by username across social media services.
Detectraptor is a repository to share detection content and queries for Velociraptor.
A researcher used ChatGPT-4o to quickly create a fake passport realistic enough to fool automated KYC systems. This raises concerns about the security of relying solely on photo-based verification methods. Experts suggest upgrading to digitally verified identity solutions to combat the growing threat of identity theft.
Microsoft has credited EncryptHub for uncovering Windows security flaws, highlighting a conflicted figure in cybersecurity and cybercrime. EncryptHub, or SkorikARI, was linked to multiple malware campaigns against high-value targets. Despite its technical skill, EncryptHub's inadequate security practices ultimately led to exposure and arrest during the Russo-Ukrainian war.
China's National Internet censor just announced new rules requiring all AI-generated content to include clearly visible or audible watermarks to indicate that it is AI-generated. Users can omit the label in some circumstances, but the AI generator must warn users of the requirement and log any requests to ignore it. These regulations will go into effect on September 1.
OpenAI is experimenting with a watermark feature for its Image Generation model in ChatGPT-4o to prevent unauthorized use of images.
The security report details a recent malware attack in which North Korean threat actors used fake recruitment emails from Dev.to to distribute BeaverTail and Tropidoor malware, which steals information, downloads additional payloads, and provides backdoor access.
International cybersecurity agencies, including the NSA and FBI, warn about the national security threat posed by Fast Flux, a technique used to hide malicious server locations.