MLflow on DanSec Daily

Cybersecurity Brief – 2026-08-21

Fri, 21 Aug 2026 00:00:00 +0000

Major Incidents or Breaches

A database belonging to the people finder service ClarityCheck was found exposed, leaking 9 million images of individuals’ faces ^[57].
Operation CameraSwarm compromised 14,000 Dahua IP cameras across Ukraine and Russia, with a focus on Russian and CIS telecom networks ^[46].
The Grandoreiro banking trojan has resurfaced in a new campaign targeting Mexico, following a previous law enforcement takedown ^[36].
A new Android malware named Manic is actively targeting Ukrainian banks, government, and identity services, as well as Russian and European financial institutions. It can exfiltrate data from offline devices by leveraging nearby infected devices ^[19]^[28].
A critical vulnerability in the Elementor Pro WordPress plugin enables attackers to upload executable files for remote code execution on affected servers ^[24].
A set of 40 malicious Mozilla Firefox extensions, posing as Web3 wallet products, have been identified stealing cryptocurrency wallet secrets ^[22].
A researcher discovered new Android malware distributed via legitimate DoFun head unit software, serving ads and building a proxy botnet ^[1].
A new report indicates that one in four data breaches now involve AI-enabled attacks, marking a 56 percent increase from the previous year ^[3].

Newly Discovered Vulnerabilities

GitLab vulnerability CVE-2026-19478 (CVSS 9.4) is under active exploitation, allowing attackers to exploit affected instances shortly after disclosure ^[6].
Microsoft Entra ID is affected by a critical remote code execution vulnerability (CVSS 10.0) that has been exploited in the wild ^[7].
Malicious versions of three Rust crates, including arrayref, were published after a maintainer account compromise, introducing build-time malware that executed on developers’ systems ^[8]^[23].
A critical vulnerability in the MLflow open-source AI engineering platform is being exploited to steal cloud credentials by allowing HTTP requests to internal endpoints ^[27]^[48].
Citrix NetScaler ADC and NetScaler Gateway are affected by two security flaws, including a critical authentication bypass. Exploitation is expected and admins are urged to patch immediately ^[14]^[26]^[52].
Zimbra Collaboration Suite (ZCS) is being actively exploited via a now-patched critical remote code execution vulnerability ^[15]^[29]^[44].
Atlassian and Splunk have issued patches for dozens of critical and high-severity vulnerabilities that could enable arbitrary code execution, privilege escalation, and data access ^[47].
Cisco has patched critical vulnerabilities in Crosswork and Secure Workload products, which could result in remote code execution, authentication bypass, and path traversal ^[49].
A critical flaw in the isolated-vm open-source JavaScript sandbox allows attackers to escape the sandbox and potentially achieve remote code execution ^[13].
NASA’s AIT-GUI operator console contains a chain of flaws allowing unauthenticated attackers to issue spacecraft commands ^[20].

Notable Threat Actor Activity

Suspected Russian cyber espionage groups have been observed abusing legitimate authentication flows, including Google OAuth and WhatsApp linking, to hijack accounts of individuals in academia, aerospace and defense, government, and NGOs ^[9].
The Head Mare hacktivist group has exploited vulnerabilities in TrueConf software to deploy PhantomCore malware ^[43].
Pakistan-linked threat actor Transparent Tribe has refreshed its toolset for cyberattacks targeting Afghan organizations, with limited success against Indian government agencies ^[35].

Trends, Tools, or Tactics of Interest

AI-generated exploit scripts are being used in active threats targeting US critical infrastructure, specifically Siemens S7 PLCs ^[11].
Attackers are exploiting trusted components, such as signed drivers and legitimate applications, to evade detection and blend malware into environments ^[10]^[53].
The Manic Android malware introduces a fallback data exfiltration mechanism using proximity to other infected devices, enabling data theft even from offline phones ^[19]^[28].
The ToxicPanda 2.0 and GoldDigger Android banking malware families have expanded their capabilities to include on-device fraud and a broader set of remote commands ^[21].
Researchers have demonstrated a “Zombie Card” attack that revives expired Visa contactless cards for in-store purchases by rewriting expiration data at the point of sale ^[16].
Two new denial-of-service (DoS) attacks exploit how major CDNs translate HTTP/3 to HTTP/1.1, resulting in amplification up to 350 times ^[18].
AI is increasingly being leveraged to personalize phishing attacks, making them harder for traditional email filters to detect ^[25].
The CUSTODY framework has been released to constrain AI agents in enterprise networks, following OpenAI-related incidents ^[31].
OpenAI has overhauled model security with new features such as sandboxing, 30-minute alerting, and training pauses in response to recent security incidents ^[51].

Regulatory or Policy Developments Affecting the Security Industry

CISA has issued warnings and urged immediate patching for actively exploited vulnerabilities in MLflow and TrueConf ^[27]^[43].
Twitch has introduced an opt-out option for users who do not want their content used for Amazon AI training ^[55].

Sources

9 million images of people’s faces exposed by reverse lookup service - malwarebytes.com
Threat Actor Hacks 14,000 IP Cameras in Ukraine and Russia - securityweek.com
'Grandoreiro' Malware Resurfaces With Mexico Campaign - darkreading.com
Manic Android Malware Exfiltrates Data From Offline Phones via Nearby Infected Devices - thehackernews.com
New Manic Android malware can exfiltrate data through nearby devices - bleepingcomputer.com
Critical Elementor Pro bug exposes WordPress sites to RCE attacks - bleepingcomputer.com
40 Malicious Firefox Extensions Pose as Web3 Products to Steal Wallet Secrets - thehackernews.com
The invisible passenger in your car - securelist.com
Report: One-Quarter of Breaches Are Enabled by AI-Driven Attacks - blog.knowbe4.com
GitLab CVE-2026-19478 Comes Under Active Exploitation Within Days of Disclosure - thehackernews.com
Microsoft Entra ID Flaw (CVSS 10.0) Exploited in Wild, Allows Remote Code Execution - thehackernews.com
Rust Supply Chain Attack Puts Build-Time Malware in Crates with 245 Million Downloads - thehackernews.com
Hackers poison arrayref Rust crate to push infostealer malware - bleepingcomputer.com
CISA warns of hackers exploiting critical MLflow vulnerability - bleepingcomputer.com
MLflow Vulnerability Exploited for Cloud Credential Theft - securityweek.com
Critical NetScaler Flaw Can Bypass Authentication on Certain Gateway and AAA Servers - thehackernews.com
Citrix urges admins to patch new NetScaler flaws as soon as possible - bleepingcomputer.com
Exploitation Expected for Critical Authentication Bypass Patched in Citrix NetScaler - securityweek.com
Attackers Exploit Zimbra SNMP Flaw for Unauthenticated Remote Code Execution - thehackernews.com
Critical Zimbra RCE flaw now actively exploited in attacks - bleepingcomputer.com
Hackers Target Zimbra Servers in Active Exploitation Campaign - securityweek.com
Atlassian, Splunk Patch Dozens of Critical, High-Severity Vulnerabilities - securityweek.com
Cisco Patches Critical Crosswork, Secure Workload Vulnerabilities - securityweek.com
Isolated-vm Flaw Lets Sandboxed JavaScript Escape to Host for Potential RCE - thehackernews.com
NASA AIT-GUI Flaws Could Let Unauthenticated Attackers Issue Spacecraft Commands - thehackernews.com
Suspected Russian Hackers Abuse Google OAuth and WhatsApp Linking to Hijack Accounts - thehackernews.com
CISA Urges Immediate Patching of Exploited TrueConf Vulnerabilities - securityweek.com
Pakistan's Transparent Tribe Refreshes Toolset for Afghan Cyberattacks - darkreading.com
AI-Generated Exploit Scripts Target Siemens S7 PLCs in U.S. Critical Infrastructure - thehackernews.com
ThreatsDay: Gogs 10.0 RCE, n8n Workflow-to-RCE, $10M Reward, GLM-5.3 AI Exploit, and More - thehackernews.com
BTR Reforged: Weaponizing Defender’s Remediation Driver as a Kernel Operation Primitive - research.checkpoint.com
ToxicPanda 2.0 and GoldDigger Expand Android Banking Attacks with On-Device Fraud - thehackernews.com
Zombie Card Attack Can Revive Expired Visa Cards for Contactless Payments - thehackernews.com
CDN Tsunami Attack Abuses HTTP/3 Translation for Up to 350x DoS Amplification - thehackernews.com
How MSPs can catch phishing attacks email filters miss - bleepingcomputer.com
New CUSTODY Framework Constrains AI Agents Inside the Network - darkreading.com
OpenAI Overhauls Model Security With Sandboxing, 30-Minute Alerts, and Training Pauses - securityweek.com
Twitch wants your content for Amazon AI training. Here’s how to opt out - malwarebytes.com

Cybersecurity Brief – 2026-08-19

Wed, 19 Aug 2026 00:00:00 +0000

Major Incidents or Breaches

The CareCloud data breach impact has grown significantly, now affecting 3.7 million individuals, far higher than the initially reported 350,000. The breach exposed sensitive health-related information^[30].
Heights Finance suffered a data breach impacting at least 1.2 million individuals, with attackers stealing names, addresses, phone numbers, Social Security numbers, and financial information from a third-party platform^[37]^[42].
SafePal disclosed that an authorization flaw in an order-tracking plugin exposed the personal data of approximately 39,798 customers, including names, email and shipping addresses, phone numbers, and purchase details^[10].
The FBI reports that Medusa ransomware has breached over 500 critical infrastructure organizations in the United States since June 2021^[11].
A single attacker has used the same infrastructure to scrape data from both Salesforce and ServiceNow customer portals across multiple industries since 2025^[8].

Newly Discovered Vulnerabilities

Three vulnerabilities in Microsoft Copilot Personal could allow a single click on a crafted link to silently exfiltrate data from connected apps and other sources^[3].
Two critical vulnerabilities have been found in MLflow (an open-source AI platform) and FUXA (web-based SCADA/HMI software), which attackers are exploiting to steal cloud credentials and secrets^[4].
A high-severity Windows Task Host vulnerability is now being actively exploited by ransomware gangs, as confirmed by CISA^[16].
A critical code injection vulnerability in GitLab (CVE-2026-19478) allows unauthenticated attackers to modify or delete user data and public projects. Mitigation is complicated by a lack of technical details^[20]^[38].
A flaw in a WordPress form plugin (CVE-2026-15748) allows unauthenticated attackers to upload executable files, potentially exposing up to 300,000 sites to compromise^[36].
Apple released updates patching 27 vulnerabilities in iOS, iPadOS, and macOS Tahoe, including a serious image-processing flaw that could allow code execution^[40].

Notable Threat Actor Activity

The Clop ransomware group deployed a custom Java web shell specifically targeting PTC Windchill and FlexPLM servers, with features for decrypting credentials and enumerating file repositories^[13].
A previously undocumented Python implant framework called TWINLOOT is abusing Microsoft SharePoint and Teams to steal credentials and move laterally across networks. The framework is modular, PyArmor-hardened, and operates entirely from Microsoft’s cloud^[7]^[24].
Ransom Busters, a ransomware affiliate, claims to have hacked other ransomware groups’ servers and is contacting victims directly, offering to delete stolen data for payments up to $60,000. The actor poses as an incident-recovery service, aiming to divert ransom payments^[5]^[23].
A new typosquatting campaign targeting RubyGems users was identified, with 16 malicious packages distributing a Windows-based information stealer that exfiltrates browser credentials and cryptocurrency wallet data^[9].
Thousands of WordPress sites have been compromised in a single operation linked to the StopAndProtect ransomware family^[39].
A China-linked threat actor demonstrated near-autonomous AI-driven attack capabilities in a campaign targeting government agencies in the Asia-Pacific region, likely Taiwan^[19].

Trends, Tools, or Tactics of Interest

Microsoft observed 7.6 billion phishing emails in Q2 2026, confirming phishing as the top initial access vector, with over 2 billion threats detected monthly^[2].
Researchers demonstrated that “AI mind viruses” can propagate between AI agents via persistent prompt files, showing the risk of self-propagating payloads in AI systems^[6].
A “CoSnitch” attack manipulated Microsoft Copilot into revealing its own security architecture, highlighting risks of meta-hacking against AI services^[21].
Microsoft has begun removing the Windows Management Instrumentation Command-line (WMIC) tool from Windows 11, addressing its abuse as a living-off-the-land binary by cybercriminals^[18].
Rapid7 reports an AI-driven surge in vulnerability disclosures and exploitation speed, straining traditional patch cycles and forcing defenders to shift prioritization strategies^[33].
The TWINLOOT framework exemplifies advanced living-off-the-land tactics, operating fully from Microsoft’s cloud and leveraging modular Python implants for persistence and credential theft^[7]^[24].

Regulatory or Policy Developments Affecting the Security Industry

OpenAI has overhauled its internal safety protocols and halted several training runs after its Astra model demonstrated “critical” cyber capabilities, raising concerns over AI agent safety^[25].

Sources

CareCloud Data Breach Impact Grows to 3.7 Million Individuals - securityweek.com
Heights Finance Data Breach Impacts at Least 1.2 Million Individuals - securityweek.com
Heights Finance data breach: What customers need to know - malwarebytes.com
SafePal Hardware Wallet Maker Says Flaw Exposed Data of Nearly 40,000 Customers - thehackernews.com
CISA: Medusa ransomware hit over 500 critical infrastructure orgs - bleepingcomputer.com
One Attacker Has Scraped Both Salesforce and ServiceNow Portals Since 2025 - thehackernews.com
Microsoft Copilot Personal Flaws Could Let One Click Exfiltrate Data From Connected Apps - thehackernews.com
Attackers Exploit MLflow SSRF Flaw to Steal Cloud Credentials and Secrets - thehackernews.com
CISA: Windows Task Host flaw now exploited by ransomware gangs - bleepingcomputer.com
Critical GitLab Zero-Click Flaw Poses Mitigation Challenges - darkreading.com
GitLab Patches Critical Code Injection Vulnerability - securityweek.com
300,000 WordPress Sites Potentially Exposed to Hacking Due to Form Plugin Flaw - securityweek.com
Apple fixes another image-processing flaw that could allow code execution - malwarebytes.com
Clop created custom web shell for Windchill data theft attacks - bleepingcomputer.com
TWINLOOT Abuses SharePoint and Teams to Steal Credentials and Move Across Networks - thehackernews.com
Silent 'TwinLoot' Cyber Threat Operates Entirely From Microsoft's Cloud - darkreading.com
Ransom Busters Claims It Hacked Ransomware Servers, Asks Victims for Up to $60,000 - thehackernews.com
'Ransom Busters': Ransomware Actor Poses as Incident-Recovery Service - darkreading.com
16 Typosquatted RubyGems Packages Steal Browser Credentials and Crypto Wallets - thehackernews.com
Thousands of Hacked WordPress Sites, One Operation: Unmasking StopAndProtect - research.checkpoint.com
China-Linked Hacker Shows AI Capabilities in APAC Attack - darkreading.com
Microsoft Observed 7.6 Billion Phishing Emails in Q2 2026 - blog.knowbe4.com
AI "Mind Viruses" Can Spread Between Agents Through Persistent Prompt Files - thehackernews.com
'CoSnitch' Attack Tricked Copilot into Mapping Out Architecture - darkreading.com
Microsoft starts removing WMIC tool used by cybercriminals - bleepingcomputer.com
AI-Driven Vulnerability Surge Breaks the Traditional Patching Model - securityweek.com
OpenAI Overhauls Safety Protocols After Its AI Agents Went Rogue - wired.com