CVE-2017-10271, CVE-2020-14882, CVE-2020-2551, CVE-2026-15981, CVE-2026-18963, CVE-2026-19478, CVE-2026-19489, CVE-2026-19490, CVE-2026-21962, CVE-2026-61979, CVE-2026-73570
Domains:
123[.]42, okta[.]com, jsdelivr[.]net, abogadosrosarinos[.]com, aptisweb[.]com, avene-hebergement[.]com, https-xhamster[.]com, co[.]id, skybap[.]shop., 194[.]31, 41[.]164, 185[.]38, milicare[.]in., keyval[.]org, glazed-client[.]com, radium-client[.]com, github[.]io, cheatlib[.]xyz, meteorclients[.]com, 22qq-client[.]com, lovable[.]app, nova-client[.]com, xenoclient[.]lol, xenonclient[.]com.
IP Addresses:
207.211.214.41, 79.127.224.14, 102.91.71.83, 162.243.116.148, 84.201.6.54, 64.225.25.188
Get tomorrow's brief in your inbox
Today: Federal agencies have 48 hours to patch a critical Oracle WebLogic flaw already under active exploitation with a CVSS 10.0 severity score. Zimbra operators face a three-day deadline for a remotely exploitable bug that allows full server takeover. A sophisticated new malware family called SynkLoader may signal incoming ransomware attacks with novel password-theft tactics.
Oracle WebLogic Server Proxy Plug-in Improper Access Control (CVE-2026-21962)
CISA added CVE-2026-21962 to the Known Exploited Vulnerabilities catalog citing active exploitation. The vulnerability allows unauthenticated attackers with network access via HTTP to compromise Oracle HTTP Server and WebLogic Server Proxy Plug-in instances, leading to unauthorized data access or modification. Oracle patched the flaw in January 2026, but exploitation attempts have continued from multiple IP addresses targeting the vulnerability alongside other persistent WebLogic RCE flaws including CVE-2020-14882/14883, CVE-2020-2551, and CVE-2017-10271. EPSS scores for these flaws range from 93.2% to 100th percentile, with three flagged as CISA-KEV entries previously linked to ransomware campaigns.
Zimbra Collaboration Suite SNMP Command Injection (CVE-2026-73570)
CISA issued a three-day remediation deadline for CVE-2026-73570, an unauthenticated remote code execution vulnerability in Zimbra Collaboration Suite. The flaw stems from improper sanitization during SNMP notification processing, allowing attackers to execute arbitrary OS commands as the Zimbra user by sending crafted SMTP requests. The vulnerability affects servers with SNMP notifications enabled, a default configuration in vulnerable versions. Poland's CERT Polska warned of active exploitation on August 16, four days before CISA's KEV addition. Exploitation gives attackers access to messages, calendars, contacts, and attachments, revealing organizational structure, administrators, vendors, naming conventions, and security processes.
GitLab Unauthenticated Code Injection (CVE-2026-19478)
GitLab released out-of-band fixes for CVE-2026-19478, a critical unauthenticated code injection vulnerability rated CVSS 9.4 affecting self-managed Community and Enterprise editions. The flaw allows remote attackers to alter or delete public projects and user data without credentials or user interaction. Exploitation attempts were observed within days of public disclosure, demonstrating the shrinking window between patch release and active exploitation.
SynkLoader Malware May Herald Ransomware Attacks
A sophisticated new malware family named SynkLoader emerged in late July 2026, using novel tactics that may foreshadow more successful ransomware attacks. The malware arrives via phishing emails from attackers who registered their own Microsoft 365 tenant using the onmicrosoft.com domain structure to impersonate victims' IT Service Desks. The installer is hosted on official Microsoft Azure storage pages to abuse brand trust. SynkLoader's technical approach is notable: it executes PowerShell instructions in temporary memory, includes a minimalist Python environment, and uses fake Microsoft runtime DLLs. The Python components reduce EDR detection signals while native DLL components enable behaviors impossible in pure Python. The malware includes a system profiler that measures victim network size and Active Directory domain membership, behaviors typically associated with ransomware groups rather than APTs. An out-of-place Python executable in a random folder under app data serves as a detection signal.
ShinyHunters vs ReliaQuest Claim Disputed
ShinyHunters added ReliaQuest to its leak site but provided no substantive proof beyond screenshots showing access to a user account on reliaquest.okta[.]com/enduser/settings. ReliaQuest published a blog post denying the breach claim, stating ShinyHunters provided no real proof because they did not gain meaningful access. The incident highlights ongoing extortion attempts by threat actors who leverage minimal access or fabricated claims to pressure victims.
Mirage2FA Campaign Hits 4,500 Organizations
The Mirage2FA phishing-as-a-service campaign targeted Microsoft 365 accounts at 4,532 unique organizations between 2024 and 2026, with 48% of targeted email addresses potentially compromised. The campaign abuses legitimate Microsoft 365 login flows to bypass two-factor authentication by stealing passwords and session cookies. The United States accounts for 63.7% of victims, with additional activity in India, Singapore, UK, Canada, and Saudi Arabia. Technology, manufacturing, and education sectors were most heavily targeted. Researchers identified more than 9,000 potential compromise events involving cookie and password theft, SSO logins, and 2FA bypass. Session theft extends beyond the initial account through SSO-connected applications and internal workflows, increasing containment costs. Unlike traditional credential phishing, attackers gain access to corporate environments through hijacked user sessions, making swift containment difficult.
WordlistLoader and SynkLoader Deliver Amatera Stealer via ClickFix
Two new malware families, WordlistLoader and SynkLoader, are being used to deliver next-stage payloads and likely sell access to ransomware groups. WordlistLoader delivers Amatera Stealer through ClearFake campaigns using ClickFix social engineering that tricks victims into running malicious commands disguised as CAPTCHA verification. Attackers compromise legitimate websites and inject Base64-encoded JavaScript blobs that fetch additional code from smart contracts stored on the blockchain (EtherHiding technique). Recent campaigns revamped to use cdn.jsdelivr[.]net to host malicious JavaScript, abusing the legitimate CDN. The ClickFix command uses conhost to launch hidden cmd.exe processes, maps remote WebDAV shares using pushd, and launches the loader via rundll32.exe. Advanced variants use conhost.exe with --headless flag to suppress console windows and employ environment variable obfuscation to conceal pushd, rundll32, and remote host names.
E4del and PINHOLE RATs Use FTP Banners as Dead Drop Resolvers
Threat actors are using FTP banners as dead drop resolvers to deliver two previously unreported RATs: E4del and PINHOLE. The technique allows malware stagers to fetch commands directly from FTP protocol initial responses, marking the first observed in-the-wild use of this method. E4del is a Node.js-based RAT embedded within a digitally signed Electron application masquerading as Discord. It supports defense evasion, persistence, system fingerprinting, encrypted C2 communication, interactive reverse shell, screenshot capture, live desktop streaming, file download, and additional payload delivery. The RAT implements a tiered jitter system with three dynamic states: Active mode (200ms-2s beaconing within first 20 seconds), Semi-Active mode (2-5s beaconing between 20-40 seconds), and Inactive mode (5-9s beaconing after 40 seconds). PINHOLE uses high-reputation platforms like Pinterest and SurveyMonkey as DDRs to obtain C2 server details and proxies communication through Cloudflare Workers.
U.S. Warns of AI-Powered Attacks on Siemens PLCs
U.S. government agencies warned that threat actors are using AI to write exploit scripts targeting internet-exposed Siemens S7 Series programmable logic controllers used across water, energy, manufacturing, and critical infrastructure sectors. This is an active threat, not theoretical risk. Attackers use legitimate scanning services like Censys and ZoomEye to identify internet-exposed or insufficiently segmented Siemens S7 PLCs. AI-generated scripts masquerade as legitimate monitoring tools to find exploits. Actors test and refine exploitation techniques against specific PLC models to improve compromise capabilities. Attackers leverage read access to understand target environments, enabling preparation and positioning for future write operations to cause disruption or operational impacts. Exploitation could result in disruption of critical industrial processes, safety incidents, downtime, equipment damage, compromise of sensitive data, compliance violations, and cascading impacts across interconnected systems.
Keycloak Password Reset Bypass Allows Account Takeover (CVE-2026-18963)
Red Hat and the Keycloak project released patches for CVE-2026-18963, a critical flaw rated 9.1 CVSS allowing unauthenticated remote attackers to take over any user account, including administrative accounts, by forcing a password reset. The vulnerability stems from improper state validation within the reset-credentials authentication flow. An attacker sends a specially crafted request to the reset-credentials endpoint, causing the authentication session to transition directly to the password update phase without requiring the email-sent action token. Successful exploitation results in complete account takeover of any user, including administrative accounts. Fixed versions: Red Hat build of Keycloak 26.4.15, 26.6.6; upstream Keycloak 26.7.2. There is no evidence of active exploitation as of August 24, 2026.
Latvia Road Traffic Safety Directorate Breach Affects 1.2 Million
Latvia's Road Traffic Safety Directorate (CSDD) confirmed a breach affecting payment records of more than 1.2 million people (roughly two-thirds of Latvia's population) and 200,000 organizations. Stolen data included identification numbers, license plates, payment amounts, dates, and addresses. Attackers exploited a vulnerability in an internet-facing system.
Sakura Internet Breach Exposes 1.36 Million Customer Accounts
Sakura Internet, a Japanese cloud and hosting provider, disclosed unauthorized access involving rental server environments and a separate sales management system. Up to 1.36 million customer accounts may have been exposed. Attackers also accessed hundreds of rental server accounts and installed malware on affected environments.
Apollo Global Data Breach via Social Engineering
Private equity giant Apollo Global Management disclosed a data breach exposing names, contact information, and SSNs following a social engineering attack that enabled threat actors to access cloud platforms between July 6-10. Apollo appears to be a victim of the UNC6671/BlackFile vishing campaign targeting private equity, financial services, and professional services sectors using IT helpdesk-themed attacks. The BlackFile group received over $10 million in Bitcoin ransom payments between January and May 2026. Multiple organizations including Blackstone, Bain Capital, KKR, TPG, Bridgewater Associates, Point72, Citadel, Two Sigma, and Millennium Management were targeted based on observed phishing infrastructure.
Berlin Government Ministries Isolated After Breach
Berlin authorities isolated the city's urban development and mobility ministries from government IT networks following a security breach. The measure disrupted email and internet access, forcing employees to use alternative communication channels and delaying several public services while ministries remained disconnected.
AI-Enabled Malware Mostly Proof-of-Concept
Palo Alto Networks Unit 42 analyzed over 400 malware samples integrating AI and found approximately 97% exist only in sandboxes and on VirusTotal, never reaching production environments. Only 12 samples appeared in Cortex XDR-protected endpoint telemetry, all detected and blocked. The AI component changes how code is authored, not how it executes, allowing existing behavioral detection, cloud-based sandboxing, and endpoint analytics to catch these threats using the same mechanisms that stop conventional malware.
Autonomous AI Agent Exploits GitHub Actions Flaw in Snowflake Repo
Researchers demonstrated an autonomous AI agent exploiting a GitHub Actions flaw in Snowflake's public repository, gaining read access to the company's internal Jira system and exfiltrating tokens within seconds with no human steering. Snowflake patched the workflow and rotated credentials after the demonstration.
14 Trojanized npm Packages Drop RedC2 4.0 Linux Backdoor
Fourteen trojanized npm packages masquerading as calendar and streak utilities were found delivering RedC2 4.0, an AI-powered Linux implant marketed on cybercrime forums as a cross-platform toolkit offering surveillance, credential theft, payload loading, and mass-operation capabilities. The version was advertised by threat actor "MarlboroMan" on Hack Forums in early June 2026.
24 npm Packages Abuse unpkg Mirrors for ClickFix Phishing
Researchers disclosed 24 npm packages used as free phishing infrastructure for redirecting to ClickFix-style fake CAPTCHA pages. Attackers use npm mirrors like unpkg to host malicious HTML files that become live, fully-rendered fake Cloudflare CAPTCHA pages on trusted domains. The HTML pages redirect to external attacker-controlled websites, with recent variants using KeyVal (api.keyval[.]org), a free public key-value store, as a dead drop resolver to extract and decode victim redirect URLs. Currently the remote logic transfers users to the legitimate ChatGPT website but could be weaponized to deliver ClickFix or other phishing domains.
Weedhack Malware Spreads via Fake Minecraft Clients
McAfee Labs detected and blocked more than 6,300 attempts to access malicious sites distributing Weedhack malware to gamers via fake Minecraft clients. Attackers create lookalike gaming websites using SEO poisoning to outrank legitimate sources in Google, Bing, Brave Search, and DuckDuckGo results. One site was built using Lovable, an AI-powered website builder, highlighting how readily available tools lower the barrier for launching convincing malicious sites. Nearly half of malicious URLs were Discord links (49.6%), followed by MediaFire (23.4%) and GitHub (8.2%).
First Malware Built for Car Head Units Fuels Botnet
Kaspersky discovered the first malware specifically designed for car head units on Android-powered aftermarket infotainment systems made by Chinese company DoFun. Attackers exploited a vulnerability in the software update distribution channel to deliver malicious Android applications serving as droppers, loaders, clickers, and reverse-proxy loaders. The malware supports nine commands primarily used to download a reverse proxy module, suggesting the main goal is enrolling devices in a proxy botnet. The malware is strongly linked to MoYu Group, one of several entities previously associated with the BadBox botnet, which has ensnared more than 10 million Android devices for fraud schemes.
Taiwan Charges 9 Over Illegal AI Server Exports to China
Taiwanese prosecutors charged nine people, including one from Nvidia and two from Super Micro, for illegally exporting high-end AI servers containing B300 graphics processing units to mainland China. Seventy-four servers arrived in China successfully (50 via Indonesia, 16 direct to China, 8 via Japan then Hong Kong). Another attempt involving 56 servers failed and the servers remain in Taiwan. Some defendants set up a company in Japan to facilitate transfers and created fake websites with falsified information to evade export restrictions.
Cisco Crosswork and Secure Workload Critical Flaws
Cisco released fixes for nine critical vulnerabilities affecting Crosswork platforms and Secure Workload software, including six flaws rated CVSS 10.0. The issues include authentication, access-control, and file-system weaknesses enabling unauthorized access or system compromise.
Citrix NetScaler ADC and Gateway Vulnerabilities
Citrix published patches for CVE-2026-19489 and CVE-2026-19490 affecting NetScaler ADC and Gateway. CVE-2026-19489 is a critical authentication bypass allowing unauthenticated attackers to access appliances configured with SAML authentication. CVE-2026-19490 causes denial of service.
miniOrange SAML Plugin WordPress Vulnerabilities Under Attack
Attackers are targeting CVE-2026-61979 (CVSS 8.1) and CVE-2026-15981 (CVSS 9.8), two severe authentication bypasses in the Xecurify miniOrange SAML 2.0 Single Sign On plugin allowing attackers to sign in as any WordPress user including administrators. CVE-2026-61979 stems from signature algorithm confusion. CVE-2026-15981 allows authentication bypass by accepting malformed signatures as valid due to loose boolean checks on openssl_verify() return values. The DigitalOcean security team discovered the vulnerabilities after observing an anomalous WordPress administrator session attempt from outside their trusted network. Scanning activity recorded from IPs: 207.211.214.41, 79.127.224.14, 102.91.71.83, 162.243.116.148, 84.201.6.54, 64.225.25.188.
NASA/JPL AIT-GUI Critical Command Execution Flaw
NASA/JPL fixed a critical vulnerability (CVSS 9.4) in the open-source AMMOS Instrument Toolkit AIT-GUI enabling unauthenticated command execution through its web console. The flaw allows remote command execution, script launches, and sequence execution.
The shrinking window between vulnerability disclosure and active exploitation is forcing organizations into emergency patch cycles. CVE-2026-73570 (Zimbra) moved from disclosure to CISA KEV listing in under two months, while CVE-2026-19478 (GitLab) saw exploitation attempts within days of public disclosure. AI is lowering barriers to exploit development, as demonstrated by attacks on Siemens PLCs using AI-generated scripts. Meanwhile, attackers are innovating with unusual C2 channels like FTP banners and npm mirror abuse to evade detection, while defenders face a "safety penalty" when frontier AI models refuse legitimate security analysis requests that attackers executing with unrestricted models never encounter.