All news with #crowdstrike tag

🔒 Since late 2024 CrowdStrike's Counter Adversary Operations has tracked MURKY PANDA, a China‑nexus actor targeting government, technology, academic, legal and professional services in North America. The group exploits internet‑facing appliances, rapidly weaponizes n‑day and zero‑day flaws, and deploys web shells (including Neo‑reGeorg) and the Golang RAT CloudedHope. CrowdStrike recommends auditing Entra ID service principals and activity, enabling Microsoft Graph logging, hunting for anomalous service principal sign‑ins, prioritizing patching of cloud and edge devices, and leveraging Falcon detection and SIEM capabilities.

🔒 Between June and August 2025, the CrowdStrike Falcon platform blocked a widespread malware campaign that attempted to compromise more than 300 customer environments. The campaign, operated by COOKIE SPIDER and renting the SHAMOS stealer (an AMOS variant), used malvertising and malicious one-line install commands to bypass Gatekeeper and drop a Mach-O executable. Falcon detections—machine learning, IOA behavior rules and threat prevention—prevented SHAMOS at download, execution and exfiltration stages. CrowdStrike published hunting queries, mitigation guidance and IOCs including domains, a spoofed GitHub repo and multiple script and Mach-O hashes.

🔒 CrowdStrike has been named the only Leader and Outperformer in the 2025 GigaOm Radar for SaaS Security Posture Management (SSPM). The recognition highlights the CrowdStrike Falcon platform's unified, AI-native approach—combining Falcon Shield, identity protection and cloud security—to detect and remediate misconfigurations, identity threats, and unauthorized SaaS access. Falcon Shield's extensive integrations, automated policy responses via Falcon Fusion SOAR, and GenAI-focused controls underpin its market-leading posture and support continuous visibility across human and non-human identities.

🔐 CrowdStrike explains how inadvertent 'leakage' — when dependent or correlated observations are included in training — can inflate machine learning performance and undermine threat detection. The article shows that blocked or grouped data splits and blocked cross-validation produce more realistic performance estimates than random splits. It also highlights trade-offs, such as reduced predictor-space coverage and potential underfitting, and recommends careful partitioning and continuous evaluation to improve cybersecurity ML outcomes.