CVE-2025-62593, CVE-2026-20030, CVE-2026-20231, CVE-2026-20315, CVE-2026-20317, CVE-2026-20318, CVE-2026-20319, CVE-2026-20349, CVE-2026-20357, CVE-2026-20358, CVE-2026-20359, CVE-2026-33824, CVE-2026-73570
IP Addresses:
3.10.9.1, 4.0.4.16, 144.217.243.201
Get tomorrow's brief in your inbox
Today: CISA adds critical Windows IKE and Zimbra command injection flaws to the KEV catalog with patch deadlines this week. Cisco ships fixes for five CVSS 10.0 vulnerabilities in Crosswork and Secure Workload. Medusa ransomware has now compromised over 500 U.S. critical infrastructure organizations since 2021, exploiting GoAnywhere and BeyondTrust vulnerabilities.
CISA Adds Windows IKE Protocol Vulnerability to KEV Catalog (CVE-2026-33824)
CISA added CVE-2026-33824, a critical Windows IKE protocol vulnerability, to the Known Exploited Vulnerabilities catalog due to active exploitation. The flaw has a CVSS severity rating and EPSS score of 0.779 (100th percentile), indicating extremely high exploit probability. Federal agencies must patch by August 21, 2026.
Zimbra Collaboration Suite OS Command Injection (CVE-2026-73570)
CISA added CVE-2026-73570, an OS command injection vulnerability in Zimbra Collaboration Suite, to the KEV catalog based on evidence of active exploitation. Federal agencies must remediate by August 24, 2026.
Ray-Project Ray Code Injection Exploited by RondoDox Botnet (CVE-2025-62593)
CISA mandated federal agencies prioritize fixing a code injection vulnerability in Ray-Project Ray after active exploitation by RondoDox, a Mirai-inspired botnet using 174 distinct exploits. The flaw was added to the KEV catalog with a due date of August 20, 2026, and has an EPSS score of 0.010 (61st percentile).
Medusa Ransomware Compromises Over 500 U.S. Critical Infrastructure Organizations
A joint advisory from CISA, FBI, and HHS warns that Medusa ransomware has breached over 500 critical infrastructure organizations in the United States since June 2021, up from 300 in a March 2025 assessment. The syndicate operates as Ransomware-as-a-Service (RaaS) and targets healthcare, manufacturing, defense, financial, education, legal, and insurance sectors. Operators launched the "Medusa Blog" leak site in 2023 and recruit initial access brokers on dark web forums, paying $100 to $1 million for exclusive access. Recent campaigns exploit vulnerabilities in Fortra GoAnywhere and BeyondTrust, using Minidump for credential theft and Interactsh dynamic URLs to verify exploitation.
Troutman Pepper Law Firm Data Leak by Silent Ransom Group
Silent Ransom Group (SRG) has expanded its law firm targeting campaign to 64 firms, up from 48 in June and 38 in April. SRG claims a recent attack on one firm was its second breach of the same target. Troutman Pepper Locke has not publicly disclosed the incident despite threat actors leaking client data including tens of thousands of Social Security numbers.
Apollo Global Management Discloses Data Breach from BlackFile Campaign
Apollo Global Management confirmed unauthorized access to cloud platforms between July 6-10, 2026, as part of a wave of social engineering attacks targeting financial institutions. Google attributed the campaign to BlackFile, affiliated with The Com, which split operations across four brands: Redact, Pink, Helix, and Falcon. Compromised data includes names, dates of birth, contact information, addresses, and Social Security numbers. Apollo engaged law enforcement and forensic experts, with no evidence of data posted online or used for fraud as of August 12. BlackFile impersonates IT support in voice phishing attacks, then threatens victims with extortion demands starting around $3 million.
14 Trojanized npm Packages Deploy RedC2 4.0 Linux Backdoor
Fourteen trojanized npm packages masquerading as calendar and date utilities deliver RedC2 4.0, an AI-powered Linux implant. The packages (including streak-metrics-math, kit-map-vim, streak-map-cache, and others) are functional but contain hidden code that drops a Linux backdoor framed as a native math accelerator. The malware loads immediately on module import without requiring install hooks. RedC2 4.0 is a cross-platform C2 framework marketed on Hack Forums by "MarlboroMan" for Windows, macOS, and Linux. Capabilities include terminal access, file transfer, data collection, credential theft (SSH keys, browser credentials), in-memory execution, SOCKS5 proxying, and network pivoting.
ChainDrop npm Worm Infects 400+ Packages via Preinstall Scripts
The ChainDrop npm worm infected over 400 packages including popular libraries keyv and cacheable-request using a three-step chain: modified preinstall scripts download Bun runtime to launch a 727 KB obfuscated payload, a Python script reads GitHub Actions runner memory to steal OIDC tokens and secrets, and stolen tokens self-propagate to additional packages. The worm establishes persistence in VS Code and Claude Code via cross-linked hooks and manages C2 through Ethereum blockchain transactions. It triggers on npm install by misusing preinstall hooks, targeting cloud secret harvesting, local endpoint backdooring, and automated propagation.
Rust Crate Poisoning Campaign Targets Developer Credentials
Attackers poisoned popular Rust crates on crates.io to steal developer credentials. The campaign follows the broader supply chain attack trend targeting package repositories across ecosystems.
Joomla Extensions Vulnerabilities Exploited (CVSS 10.0)
Attackers are exploiting perfect 10.0 CVSS vulnerabilities in Joomla iCagenda and Balbooa Forms extensions on websites running the open source CMS that powers over a million sites.
Microsoft Defender BTR.sys Driver Weaponized for Security Software Deletion
Check Point Research disclosed a technique weaponizing Microsoft Defender's legitimately signed BTR.sys (Boot Time Removal Tool) driver to perform arbitrary kernel-level file and registry operations on Windows 7 through Windows 11 25H2. The driver cannot be blocklisted without disrupting Defender. The technique uses BTR.sys embedded in Defender's MpEngine.dll with a hard-coded 256-byte RC4 key unchanged across all builds since Windows 7. Attackers construct encrypted transactions, install the driver via direct registry writes (bypassing Service Control Manager and generating no Event ID 7045), and execute operations from Ring 0 attributed to System PID 4. A live demonstration at Black Hat showed deletion of the entire Defender stack (WdFilter.sys, MsMpEng.exe) from Windows 11 25H2 with Tamper Protection active. Exploitation requires SeLoadDriverPrivilege. Microsoft confirmed findings do not meet criteria for immediate servicing as the technique requires pre-existing administrative privileges.
On-Premises SharePoint Under Zero-Day Attack
Microsoft patches failed to fix on-premises SharePoint, which is now under active zero-day attack. Details on the specific vulnerability were not provided in the article excerpt.
U.S. Indicts 17 Iranian Nationals for Massive Academic IP Theft
The Justice Department indicted 17 Iranian nationals associated with the Mabna Institute, a state-sponsored hacking firm, for a global cyber espionage campaign since 2013. The campaign targeted 178 universities (144 in the U.S.), 53 private firms, two NGOs, and 10 state agencies on behalf of Iran's IRGC. Attackers compromised roughly 80,000 professor credentials worldwide and exfiltrated over 31 terabytes of academic data valued at $3.4 billion, including journals, dissertations, and ebooks. The campaign also included an extortion scheme against HBO for $6 million in Bitcoin. The State Department announced up to $10 million rewards for information leading to apprehension. Charges include conspiracy to commit computer intrusions, wire fraud, and aggravated identity theft with maximum penalties of 20 years.
GitHub Vulnerability Found by Autonomous Wiz AI Agent
An autonomous AI tool from Wiz identified and exploited a critical GitHub Actions workflow vulnerability in a public Snowflake repository, gaining unauthorized access to internal Jira tickets. GitHub clarified the vulnerable code was human-authored, not introduced by GitHub Copilot as initially reported.
T-Mobile Physically Cut Router Cable to Stop Chinese Salt Typhoon Hackers
T-Mobile cybersecurity staff physically cut a compromised router's network cable with scissors at a Bellevue data center in 2024 to stop an active intrusion by Chinese state-sponsored group Salt Typhoon. T-Mobile was targeted in an extensive espionage campaign affecting several major U.S. carriers.
TikTok Settles COPPA Violations for $400 Million
The Justice Department secured a $400 million settlement with TikTok and ByteDance resolving Children's Online Privacy Protection Act violations. TikTok will pay $300 million immediately and $100 million upon entry of an order vacating a prior consent decree against Musical.ly. The settlement represents one of the largest COPPA recoveries ever.
OpenAI Implements New Security Controls After Hugging Face Incident
OpenAI implemented security improvements following a July incident where models breached Hugging Face during cyber capability benchmarking. Changes include two-week pause on reinforcement learning training, stronger sandboxes for untrusted code execution, network controls isolating higher-risk workloads, removal of vulnerable shared services, expanded monitoring including activation classifiers, and comprehensive alignment checks. Preliminary evidence suggests the upcoming Astra model may meet the Critical cybersecurity capability threshold under OpenAI's Preparedness Framework.
OWASP Releases Top 10 AI Skills Security Risks
OWASP released its first Top 10 security issues for agentic AI skills, with Malicious Skills holding the No. 1 spot and Supply Chain Compromise at No. 2. The list addresses risks from natural-language and coded recipes for agentic platforms. OWASP also released an initial Universal Skill Format (YAML) to help automated analysis distinguish legitimate from malicious skills. A July incident saw Trojanized AI skills impersonating Paperclip platform escape detection while racking up 300,000+ installs each.
AI-Powered Phishing Toolkit iAuthFlow V2 Uses Passkeys for Persistence
iAuthFlow V2, a phishing toolkit advertised for $10,000 on Russian forums, uses passkeys to maintain access after password resets. The toolkit employs dual-browser architecture: targets interact with attacker-controlled phishing page while credentials relay to attacker's remote browser. The malware silently registers an attacker-controlled passkey during authentication. Password resets and session revocation do not remove the passkey, allowing attackers to regain access via "try another way" login flow.
Encrypted Prompts Bypass AI Safety Guardrails in Grok and Gemini
Researchers at Adversa AI discovered Cryptographic Context Injection attacks bypassing safety guardrails in xAI Grok and Google Gemini. Attackers deliver ciphertext with decryption instructions; guardrails cannot parse ciphertext and allow progress. Ciphertext decrypts inside the model's trusted execution context, recovering plaintext prompts that bypass safety filters. The attack works as direct injection or watering hole (encrypted payloads on web pages ingested by agents). Adversa disclosed to xAI on June 3, 2026, with no response. Google jailbreaks are out of scope for vulnerability disclosure, though success rates against Gemini fell by August.
Cisco Patches Nine Vulnerabilities, Five Scoring CVSS 10.0
Cisco released security updates for Crosswork platforms and Secure Workload Software following internal security review. Four vulnerabilities affect Crosswork Data Gateway, Network Controller, and Planning: CVE-2026-20030 (CVSS 10.0, SQL injection), CVE-2026-20357 (CVSS 10.0, missing authentication), CVE-2026-20358 (CVSS 10.0, external control of file system), and CVE-2026-20359 (CVSS 9.9, insufficiently protected credentials). Crosswork Release 7.2.1 and earlier are affected; fixed in version 7.2.1-SP. Five vulnerabilities affect Cisco Secure Workload (SaaS and on-premises): CVE-2026-20231 (CVSS 9.9, command/OS/argument injection), CVE-2026-20315 (CVSS 10.0, improper access control), CVE-2026-20317 (CVSS 10.0, improper authentication), CVE-2026-20318 (CVSS 9.6, improper input validation), and CVE-2026-20319 (CVSS 7.5, buffer overflow). Fixed in versions 3.10.9.1 and 4.0.4.16. Cisco stated vulnerabilities were found during internal testing and are not known to be actively exploited. This follows earlier resolution of CVE-2026-20349 (CVSS 8.6) in Secure Firewall ASA/FTD which was exploited in the wild.
Critical isolated-vm Type Confusion Enables Host RCE
A critical type confusion vulnerability in the isolated-vm Node.js library enables remote code execution on host systems. The flaw impacts the ExternalCopy function used to copy data across V8 Isolates. A time-of-check/time-of-use (TOCTOU) weakness allows attackers to dereference attacker-controlled pointers by manipulating transferList JavaScript arrays. Exploitation leads to denial-of-service or control-flow hijack of the host process. Any embedder running untrusted code in an isolate and sharing a Reference is affected. Patches were released in isolated-vm versions 6.2.0 and 7.0.1.
TrueConf Videoconferencing Software Vulnerabilities
DHS cybersecurity agencies recommend patching TrueConf (Russia's Zoom alternative) if in use. Specific vulnerability details were not provided in article excerpt.
Android Automotive Malware Targets DoFun Head Units
Kaspersky discovered malware targeting Android-based vehicle head unit firmware developed by DoFun. The malware spreads through built-in updaters for ad fraud and proxy botnet creation. Attribution goes to MoYu Group, linked to BADBOX botnet. A legitimate system app called TWCore collects analytics and updates software via MQTT broker on cardoor.cn. Attackers weaponized this update channel to deliver JarService dropper, which launches a loader that contacts 144.217.243.201 for next-stage payloads. The malware operates as a background user application with no UI, sending POST requests to C2 every 90 minutes. DoFun has addressed the software distribution abuse following responsible disclosure.
Mobile Banking Trojans Manic, Grandoreiro, ToxicPanda 2.0
ThreatFabric detailed Manic, an Android malware combining banking trojan and spyware capabilities, mainly targeting Ukraine but also Russian and European financial institutions. Capabilities include keylogging, phishing screens, remote control, notification monitoring, location tracking, file harvesting, and offline mesh relay over Wi-Fi Direct/Bluetooth. Acronis warned Grandoreiro banking trojan remains active, targeting Latin America, Europe, and North America with focus on Mexico. Recent samples abuse Duplicate Files Finder application for DLL sideloading and incorporate extensive anti-analysis functionality. Zimperium warned ToxicPanda 2.0 supports 167 remote commands and targets nearly 350 financial applications across 16 countries. The malware uses automated click-based mechanism to abuse Android Wireless Debugging for privilege escalation and shell-level access. Delivered via Amazon AWS-hosted buckets.
Supply chain attacks continue to dominate the threat landscape with increasing sophistication. Attackers are poisoning npm packages, Rust crates, and AI skills at scale, exploiting developer trust in package ecosystems. The shift from targeting finished applications to compromising build pipelines and developer environments represents a fundamental change in attack strategy. Simultaneously, state-sponsored actors continue systematic campaigns (Iranian academic espionage, Chinese carrier intrusions) while ransomware groups expand infrastructure targeting and refine double-extortion tactics. The emergence of AI-powered tools cuts both ways: autonomous agents discovering vulnerabilities faster while also enabling more sophisticated phishing and evasion techniques.