Sponsored
2.4X
Weak and reused passwords expose companies to critical risks. In fact, the majority of web application breaches stem from poor credential hygiene, costing organizations millions per incident.
Parse notes
Sponsored
Weak and reused passwords expose companies to critical risks. In fact, the majority of web application breaches stem from poor credential hygiene, costing organizations millions per incident.
Nearly all of enterprise customers report that Bitwarden improved their security posture once fully deployed. Start a free enterprise
Over 252 million identity records from seven countries—Turkey, Egypt, Saudi Arabia, UAE, Mexico, South Africa, and Canada—were exposed through three misconfigured servers in Brazil and the UAE. These contained government-level profiles with ID numbers, birth dates, contact details, and addresses. The databases had similar structures, suggesting a single operator, but the source remains unknown. Turkey, Egypt, and South Africa were most affected. Cybernews contacted hosting providers, and the data is no longer publicly accessible as of June.
Chess.com reported that it experienced a data breach after detecting unauthorized data stored in a third-party file transfer application. The incident only impacts an estimated 4.5K users and may have exposed names and other PII.
TP-Link has confirmed the existence of an unpatched zero-day vulnerability impacting multiple routers. The vulnerability was discovered by a security researcher who stated that it is caused by a stack buffer overflow in the CPE WAN Management Protocol (CWMP). An attacker could exploit it by redirecting vulnerable devices to a malicious CWMP server, which could deliver an oversized SOAP payload, leading to remote code execution.
Attackers deployed fake GitHub repositories and npm packages that stored malware download URLs in Ethereum smart contracts rather than directly in the code, making detection significantly harder for security tools. The campaign targeted cryptocurrency developers with fake automated trading bots, using packages like colortoolsv2 and mimelib2 that connected to blockchain addresses to retrieve second-stage payloads. The attackers automated thousands of fake commits daily to inflate repository legitimacy, demonstrating sophisticated supply chain attack techniques that combine social engineering with blockchain-based evasion.
A new vulnerability in popular Electron-based apps like Signal, 1Password, and Slack allows attackers to get around code integrity checks by hiding backdoors in overlooked V8 heap snapshot files. Electron's integrity protections often miss these files, meaning backdoors can be inserted without failing code signatures or integrity checks, especially when apps are installed in user-writable folders. This post shows a proof of concept to demonstrate keylogging and privilege escalation with the exploit, showing that even patched applications may have missed attack surfaces.
Sponsored
Most SOC teams focus on reactive work but aren't actively looking for threat actors that have bypassed their security controls. TryHackMe's THREAT HUNTING SIMULATOR gives analysts a safe, repeatable environment to practice spotting adversary behaviours [12]. Gamified hunts and instant feedback enable teams to sharpen detection rules and response playbooks, without the overhead of building internal labs. Start free
Fraction Loader is an educational malware project that creates a stealthy in-memory loader capable of downloading, reassembling, and executing payload fragments without touching disk. It uses Windows APIs like VirtualAlloc and WinHTTP to handle encrypted payload chunks, parse headers for order, and reassemble them into executable memory with suitable permissions. While meant for penetration testers and security researchers to learn evasion techniques, it functions as a weaponized dropper that can bypass file-based detection and pose a threat to InfoSec professionals monitoring for in-memory attacks.
Tidal Cyber maps tactics, techniques, procedures, and procedural-level insights to MITRE ATT&CK to tailor security strategies around adversary groups and their behavior for threat-led defenses.
Mixeway Flow is a DevSecOps platform that automatically performs four security scans (SAST, SCA, IaC, and secret detection) on Git repositories through webhook integration. It aggregates vulnerabilities from various open-source scanning tools such as Bearer, OWASP Dependency Track, KICS, and Gitleaks into a single dashboard for streamlined security management. The platform simplifies CI/CD pipeline setup and offers thorough application security testing across code, dependencies, infrastructure, and credential exposure.
Egyptian authorities and the Alliance for Creativity and Entertainment (ACE) shut down Streameast, the world's largest illegal sports streaming platform that received 136 million monthly visits across 80 domains, arresting two suspects and seizing assets linked to $6.2 million in laundered advertising revenue.
A California jury found Google liable for collecting user data even after tracking was disabled, resulting in a $425 million penalty. The class action lawsuit covered 98 million users and claimed Google continued to gather app activity despite settings meant to protect privacy.
France's data authority fined Google €325M and SHEIN €150M for using cookies without user consent and breaching privacy laws. Google was also penalized for showing ads in Gmail based on those cookies. Meanwhile, SHEIN's opt-out failed as cookies were still sent even when users clicked “Reject All.”
Texas Attorney General Ken Paxton filed suit against education software provider PowerSchool for a December 2024 data breach that exposed personal information of more than 880,000 Texas students and teachers, alleging the company failed to implement adequate security measures despite knowing about vulnerabilities.
Disney agreed to pay $10 million to settle FTC claims that it mislabeled child-directed YouTube videos as "Not Made for Kids," which enabled unlawful collection of children's personal data and targeted advertising without parental consent in violation of COPPA regulations.
The FreeBSD Project is developing a policy to restrict LLM-generated code contributions due to licensing concerns, though AI may still be allowed for documentation, translations, and code analysis tasks.