Full Daily Digest

The attack chain began with the exploitation of a vulnerable SQL Server, though the exact entry point remains unclear. Once inside, attackers deployed a malicious DLL via sideloading, modified system settings, and used vulnerable drivers (e.g., Huawei's HWAuidoOs2Ec.sys) to gain kernel-level privileges and disable security tools. The novel use of Microsoft Teams' TURN infrastructure for C2 traffic complicates detection, as outbound connections appear legitimate. Post-ransomware deployment, Backdoor.Turn was injected into DbgView64.exe, enabling persistent access and lateral movement. Mitigations include monitoring for anomalous Teams traffic, restricting driver loading, and applying patches for known CVEs (e.g., CVE-2023-52271).

Help Net Security · 2026-06-16 · Incidents: DragonForce ransomware group abuses Microsoft Teams TURN relays to hide C2 traffic, evading detection for months.

The vulnerability (CVE-2026-20262) stems from insufficient input validation in Cisco Catalyst SD-WAN Manager's web UI, allowing authenticated attackers to traverse directories and overwrite files via crafted HTTP requests. This flaw affects all deployment types (on-prem, cloud, FedRAMP). Attackers leverage it to deploy malicious .war files, executed via WildFly, indicating post-compromise command execution. The identical patch list for CVE-2026-20245 suggests either concurrent remediation or a shared root cause. Federal agencies must patch by June 29, 2026, per CISA's directive. Organizations should prioritize log review for IoCs (e.g., .war file deployment) and restrict write access to critical directories.

Help Net Security · 2026-06-16 · Vulnerability: Cisco's SD-WAN Manager faces another exploited path traversal flaw (CVE-2026-20262), enabling file manipulation and root escalation, with patches overlapping a prior vulnerability.

The attack begins with a spear-phishing email impersonating Microsoft security alerts, creating urgency to open a malicious ZIP attachment. The LNK file inside triggers a batch script that downloads Python and a CAT file, enabling in-memory execution of NarwhalRAT. Persistence is achieved via scheduled tasks mimicking legitimate Microsoft processes. The malware's use of Korean websites and pCloud API for C2 communication demonstrates sophisticated evasion techniques. Defenders should scrutinize unexpected security alerts, disable LNK file execution where possible, and monitor for unusual scheduled tasks. Organizations in sectors targeted by North Korean APTs (e.g., government, defense) should prioritize user awareness training and endpoint detection for Python-based malware.

The Hacker News · 2026-06-16 · Incidents: ScarCruft deploys NarwhalRAT via fake Microsoft alerts, using multi-stage infection and cloud-based C2 channels.

The attack begins with a phishing email containing a disguised business document, which triggers an obfuscated batch file to initiate a multi-stage infection chain. Phantom Stealer harvests credentials, session cookies, and financial data from major browsers, while also capturing keystrokes, screenshots, and clipboard contents. Its MaaS model allows multiple threat actors to deploy and update the malware, increasing its reach and adaptability. Organizations must prioritize behavior-based endpoint detection and response (EDR) solutions to identify anomalous activities like abnormal command-line executions. Additionally, monitoring for unusual outbound traffic to known exfiltration channels (Telegram, Discord, FTP, SMTP) is critical.

Dark Reading · 2026-06-16 · Incidents: Phantom Stealer leverages fileless techniques to steal browser credentials and financial data, evading traditional defenses through memory-only execution and multi-channel exfiltration.

The export control on Mythos 5 and Fable 5 reflects a reactive stance to perceived risks of LLM-assisted exploit development, but fails to account for adversarial access to comparable models. Mythos's capability to autonomously discover critical vulnerabilities could democratize offensive tooling, yet defenders rely on the same technology for patch verification and threat modeling. The ban creates an asymmetric disadvantage: state-sponsored actors likely already possess similar tools through open-weight models or clandestine development. Mitigation requires calibrated access controls, such as verified defender programs, rather than blanket restrictions. Proactive collaboration between vendors and policymakers could address misuse risks without stifling defensive innovation.

Dark Reading · 2026-06-16 · Policy: Cybersecurity experts condemn US export restrictions on Anthropic's Fable 5 and Mythos 5 models, warning it cripples defenders while adversaries advance.

The administration's abrupt export controls on Anthropic's AI models reveal a fragmented approach to AI regulation, lacking clear technical or threat-based rationale. This creates uncertainty for cybersecurity professionals who rely on these tools for defense. The move risks disadvantaging U.S. companies and allies while failing to address the broader need for standardized AI security frameworks. Mitigation should include transparent threat assessments, stakeholder engagement, and clear compliance guidelines to avoid ad hoc disruptions. The administration must balance national security with innovation, ensuring decisions are evidence-based rather than politically driven.

CyberScoop · 2026-06-16 · Policy: Lawmakers question the Trump administration's export controls on Anthropic's AI models, citing unclear national security grounds and potential political motivations.

The rapid iteration of AI models like Claude Code creates a unique attack surface where vulnerabilities such as prompt injection and arbitrary code execution can be exploited before patches are widely adopted. Unlike traditional software, AI models require manual updates or environment testing, delaying critical fixes. Attackers can exploit these gaps to poison data, leak credentials, or implant backdoors. Mitigations include implementing staged rollouts, monitoring update logs for silent patches, and enforcing strict prompt validation. Organizations must balance model performance with security, treating AI tools as dynamic, evolving systems rather than static software.

CyberScoop · 2026-06-16 · Vulnerability: Rapid AI model updates introduce silent security gaps, forcing organizations to choose between stability and patched vulnerabilities.

The SprySOCKS Windows variant represents a significant evolution in FishMonger's toolkit, combining cross-platform capabilities with advanced evasion techniques. The malware's use of encrypted kernel drivers (e.g., fsdiskbit.sys) demonstrates a focus on persistence and stealth, though reliance on leaked certificates suggests opportunistic rather than highly sophisticated tradecraft. Attack paths likely involve exploiting N-day vulnerabilities or misconfigured public-facing servers, with potential UEFI bootkit involvement (CVE-2023-24932). Defenders should prioritize HVCI deployment, patch management, and monitoring for IoCs like the hardcoded C2 server IP. The campaign underscores the growing risk of kernel-level attacks in enterprise environments.

Dark Reading · 2026-06-16 · Incidents: FishMonger's SprySOCKS Windows variant abuses kernel drivers for stealth, targeting governments in Asia and Latin America.

The vulnerability exploited predictable bucket naming in Google Cloud's Vertex AI SDK, allowing attackers to pre-create buckets in their own projects. When victims uploaded models without specifying a bucket, the SDK used the predictable name, enabling attackers to replace the model with a malicious one. The attack leveraged Python's pickle serialization, which executes code upon loading. Unit 42's proof of concept demonstrated a 1.4-second window to swap the model, stealing OAuth tokens and accessing sensitive data. Mitigations include updating to SDK version 1.148.0, explicitly setting staging buckets, and auditing client-side SDK usage across all environments. This flaw underscores the risks of default configurations and the need for robust ownership checks in cloud services.

The Hacker News · 2026-06-16 · Vulnerability: Google Vertex AI SDK flaw let attackers hijack model uploads via predictable bucket names, enabling code execution in Google's infrastructure.

The ClickFix campaigns demonstrate a sophisticated evolution in malware delivery, leveraging social engineering to bypass user defenses. Attackers use fake update lures and fraudulent troubleshooting instructions to execute PowerShell commands, deploying loaders like BabaDeda, which perform host profiling and evade security products. These loaders then drop info-stealers or RATs, often via trusted processes like svchost.exe. The campaigns' success lies in their simplicity—exploiting human trust in authoritative instructions. Mitigations include user education on command execution risks, endpoint detection for PowerShell abuse, and application allowlisting to block unauthorized scripts. Organizations should also monitor for lateral movement techniques like WMIExec and SMBExec.

The Hacker News · 2026-06-16 · Incidents: ClickFix campaigns deploy new malware loaders via social engineering, targeting sectors with fake updates and PowerShell exploits.

Rokarolla's attack path begins with social engineering, luring users to malicious sites posing as legitimate app distributors. Once installed, it gains administrative privileges, enabling a broad range of malicious activities, from credential theft to remote control. The malware's persistence mechanisms, such as disabling Google Play Protect and suppressing device audio, make it particularly resilient. Affected scope includes 217 apps, primarily in banking and cryptocurrency sectors. Mitigation requires a multi-layered approach: deploying mobile threat defense solutions to detect behavioral anomalies, enforcing strict app installation policies, and replacing SMS-based authentication with phishing-resistant MFA. Organizations must treat Android devices as high-risk endpoints, given their increasing role in corporate networks.

Dark Reading · 2026-06-16 · Incidents: Rokarolla Android Trojan evolves beyond banking fraud to full device control, persistence, and extensive surveillance.

The push for AIBOMs addresses a critical gap in AI security: the lack of visibility into model components and training data. Attackers could exploit undocumented dependencies or poisoned datasets, leading to downstream vulnerabilities in AI-driven systems. The paper correctly identifies the chicken-and-egg problem in adoption, where neither suppliers nor buyers prioritize documentation without external pressure. Mitigation requires coordinated action: 1) Vendors must catalog datasets, training methods, and model architectures; 2) Procurement teams should mandate AIBOMs in contracts; 3) Regulators must define minimum disclosure standards, akin to PCI-DSS for payment systems. Without this, AI supply chains remain opaque and vulnerable to subcomponent attacks.

CyberScoop · 2026-06-16 · Policy: Standardized AI Bills of Materials (AIBOMs) are proposed to improve AI transparency and mitigate supply chain risks.

The temporary block on Telegram underscores a reactive approach to exam fraud, targeting the platform rather than addressing insider leaks in the exam logistics chain. Attackers exploited Telegram's message-editing feature to post fake questions pre-exam, later replacing them with real ones to simulate leaks. This tactic misled candidates into paying for non-existent leaks, with fraud rings moving significant sums through fraudulent accounts. Mitigation should focus on securing the exam supply chain, implementing digital forensics to trace leaks, and educating candidates on fraud tactics. Platform-level blocks are blunt instruments; targeted takedowns of fraudulent channels and criminal prosecutions would be more effective without disrupting legitimate use.

The Record by Recorded Future · 2026-06-16 · Incidents: India blocks Telegram to curb exam cheating, sparking debate over effectiveness and digital rights.

The exploitation of these FortiSandbox vulnerabilities represents a significant escalation in attacker tactics, targeting a core component of Fortinet's security infrastructure. CVE-2026-39813's path traversal flaw allows authentication bypass via the JRPC API, while CVE-2026-39808 and CVE-2026-25089 enable unauthenticated command injection through crafted HTTP requests. The vibecoded exploit for CVE-2026-25089, though potentially faulty, indicates attackers are testing exploit chains against FortiSandbox—a historically less-targeted system. This shift aligns with broader trends of AI lowering exploit development barriers. Mitigations include immediate patching, network segmentation to restrict access to FortiSandbox management interfaces, and enhanced monitoring for anomalous HTTP requests targeting these endpoints.

Help Net Security · 2026-06-16 · Vulnerability: Attackers exploit FortiSandbox vulnerabilities (CVE-2026-39813, CVE-2026-39808, CVE-2026-25089) for RCE and authentication bypass, signaling increased targeting of this critical security component.

The Lorem Ipsum campaign exemplifies modern threat actors' adaptability, pivoting from signed-installer delivery to exploiting compromised WordPress sites via ClickFix lures after losing access to fraudulent certificates. This shift not only expands the attack surface but also aligns with tactics used by Rapid Brigantine, a group known for ransomware and data extortion. The campaign's resilience underscores the importance of dynamic defense mechanisms. Organizations should prioritize behavioral analytics to detect anomalous activities across multiple delivery channels, implement robust web application security measures to prevent WordPress compromises, and educate users on recognizing social engineering lures.

Dark Reading · 2026-06-16 · Incidents: Lorem Ipsum malware now uses ClickFix lures on WordPress sites, linked to ransomware group Vice Society, requiring behavioral detection strategies.

The launch of TekStream's Proactive Cyber Defense service reflects a strategic shift from reactive to proactive security measures, driven by the rapid evolution of AI-powered threats. Autonomous attack systems, such as Anthropic’s Mythos, exemplify how adversaries leverage machine-speed exploitation, necessitating defenses that operate continuously and adaptively. The service’s integration with Cosmos ensures that intelligence compounds across threat landscapes, improving detection and hardening workflows. This model is particularly effective in fragmented environments where traditional security tools like MDR and SIEM often operate in silos. Mitigation strategies include continuous threat hunting, exposure reduction, and operational coordination, which collectively reduce exploitable attack surfaces. Organizations should prioritize integrating such proactive defenses to stay ahead of AI-accelerated threats.

Help Net Security · 2026-06-16 · Tools: TekStream's Proactive Cyber Defense service combats AI-driven threats with continuous, intelligence-driven security enhancements.

The vulnerability exploits an OIDC authentication bypass in SimpleHelp, allowing attackers to self-register Technician accounts even with MFA enforced. Successful exploitation hinges on three conditions: OIDC provider configuration, TechnicianGroup association, and enabled group authentication. Attackers can then execute scripts, remote into endpoints, and pivot across networks. With 14,000 servers exposed and 7.2% vulnerable, the risk is significant, especially for MSPs. Horizon3.ai's findings suggest ransomware groups may quickly weaponize this flaw, as seen with prior SimpleHelp vulnerabilities. Mitigations include upgrading to patched versions, disabling internet access for servers, and scrutinizing logs for unauthorized Technician activity. Network segmentation and IP restrictions can further reduce exposure.

Help Net Security · 2026-06-16 · Vulnerability: SimpleHelp RMM flaw (CVE-2026-48558) enables unauthenticated attackers to create Technician accounts and compromise endpoints via OIDC misconfiguration.

Rokarolla represents a sophisticated evolution of Android banking trojans, leveraging Accessibility permissions to bypass traditional defenses like Google Play Protect. The malware's attack path begins with a dropper masquerading as Play Protect, which installs the payload and grants itself Accessibility access. Once active, Rokarolla downloads fake HTML overlays for targeted apps, capturing login credentials and card details. It also mimics the lock screen to harvest PINs, enabling persistent device control. The malware's ability to intercept SMS codes and manipulate the clipboard for crypto theft underscores its financial motivation. Mitigation requires strict adherence to app source verification, enabling Play Protect, and scrutinizing Accessibility requests. Organizations should also monitor for IOCs shared by Zimperium.

The Hacker News · 2026-06-16 · Incidents: Rokarolla Android malware exploits Accessibility permissions to steal banking details, SMS codes, and crypto wallet funds via overlays and clipboard manipulation.

The rapid adoption of AI agents introduces a significant attack surface, as these entities operate autonomously with broad access to sensitive systems. Without proper governance, they can misuse privileges or violate compliance, exposing organizations to security risks. Concurrently, quantum computing threatens current cryptographic methods, necessitating post-quantum-ready PKI. AppViewX's solution combines AI agent governance with cryptographic depth, addressing both challenges. Mitigations include granular access control, real-time threat detection, and integration with existing security tools. This approach reduces the risk of unmonitored agent activities and future-proofs cryptographic infrastructure.

Help Net Security · 2026-06-16 · Tools: AppViewX introduces Agent Identity Security to govern AI agents and secure post-quantum cryptographic environments.

Email authentication remains a critical yet often overlooked defense against phishing and spoofing attacks. Cloudflare's DMARC Management tool addresses two key pain points: complexity and cost. The SPF lookup audit feature is particularly noteworthy, as exceeding the 10-DNS-lookup limit is a common misconfiguration that can lead to email delivery failures. Attackers exploit weak DMARC policies (e.g., p=none) to impersonate legitimate domains, bypassing traditional spam filters. Organizations should prioritize moving to p=quarantine or p=reject to mitigate these risks. The tool's integration with Cloudflare DNS simplifies deployment, but organizations using other DNS providers must migrate to benefit. Future enhancements like forensic reporting could further streamline incident response.

Cloudflare Blog · 2026-06-16 · Tools: Cloudflare's free DMARC Management tool now helps organizations achieve email authentication compliance with SPF audits and enforcement recommendations.

The lack of identity and access control in AI agent workflows poses significant risks, particularly when agents interact with production infrastructure. Attack paths could involve malicious instructions or unauthorized access to sensitive resources. Teleport's solution mitigates these risks by embedding enforcement at the inference endpoint (LLM Proxy) and dynamically assigning scoped permissions (Delegated Identity). This approach limits exposure by ensuring agents operate with minimal privileges and all actions are logged. Organizations deploying AI agents should prioritize such frameworks to prevent credential misuse and ensure auditability. Implementing these controls reduces the risk of lateral movement and data exfiltration by compromised agents.

Help Net Security · 2026-06-16 · Tools: Teleport's new LLM Proxy and Delegated Identity enforce zero trust and auditability for AI agent workflows.

The AI Xploit Shield represents a significant advancement in vulnerability management by addressing the critical gap between discovery and remediation. Traditional patching processes are often delayed due to testing requirements, legacy systems, and third-party dependencies, leaving systems vulnerable. This solution uses AI to generate and deploy tailored virtual patches in near real-time, effectively blocking exploitation attempts without modifying the underlying application. The approach is particularly relevant as emerging AI models accelerate vulnerability discovery, increasing the volume of new vulnerabilities. Organizations can now mitigate risks more effectively, buying time for thorough patch validation and deployment.

Help Net Security · 2026-06-16 · Vulnerability: Radware's AI Xploit Shield automates virtual patching to protect applications and APIs from newly discovered vulnerabilities.

The article underscores a critical blind spot in secrets management: developer workstations, where credentials persist in plaintext across shell histories, local caches, and AI agent configurations. Attackers increasingly target these endpoints via infostealers or supply chain compromises, bypassing traditional repository-focused defenses. GitGuardian's solution leverages Rust-based scanning for speed and integrates with existing workflows via ggshield, but it doesn't automate credential rotation or remediation. Organizations must now treat developer laptops as high-risk assets, correlating endpoint findings with vaults and identity systems to mitigate lateral movement. The rise of AI coding tools like Copilot accelerates credential sprawl, demanding proactive visibility before breaches occur.

GitGuardian Blog · 2026-06-16 · Tools: GitGuardian expands secrets detection to developer endpoints, addressing AI-driven credential exposure risks with a fast, privacy-first scanning tool.

The proliferation of VPNs and residential proxies has enabled attackers to obscure their origins, making IP-based detection unreliable. Attack paths often involve credential stuffing or account takeovers, where malicious traffic mimics legitimate user behavior. Affected scope includes financial losses and operational disruptions, particularly in sectors reliant on user authentication. Mitigations should focus on behavioral analytics, integrating IP intelligence into automated workflows, and adopting risk-based access controls. Security teams must prioritize attribution and infrastructure context over raw data volume to reduce false positives and investigation time.

The Hacker News · 2026-06-16 · Incidents: Anonymized infrastructure appears in 94% of incidents, yet many organizations lack effective workflows to respond.

The exploitation of these FortiSandbox vulnerabilities highlights a persistent targeting of Fortinet's security appliances. The attack path involves unauthenticated HTTP requests to the JRPC API and WEB UI, enabling command injection and authentication bypass. Affected scope includes FortiSandbox, FortiSandbox Cloud, and FortiSandbox PaaS. Mitigations include immediate patching, network segmentation to restrict access to management interfaces, and monitoring for anomalous HTTP requests. The AI-assisted exploit attempt for CVE-2026-25089, though faulty, signals a concerning trend in automated vulnerability weaponization. Defenders should prioritize patch management and consider deploying intrusion detection rules for known exploit patterns.

The Hacker News · 2026-06-16 · Vulnerability: Attackers exploit three critical FortiSandbox flaws, including a recently patched AI-assisted exploit attempt.

The attack path begins with social engineering via social media or fake investment personas, luring victims into fraudulent crypto schemes. When financial institutions block transfers, scammers pivot to in-person cash pickups, using couriers authenticated by serial numbers or passwords. This tactic exploits trust and urgency, often demanding additional fees for fictitious taxes or penalties. The scope is broad, targeting individuals globally, with losses exceeding $11 billion. Mitigations include verifying investment platforms through independent reviews, scrutinizing URLs, and recognizing 'love bombing' tactics. The FBI emphasizes vigilance against unsolicited investment offers and unexpected cash pickup requests.

Help Net Security · 2026-06-16 · Incidents: Scammers use couriers to collect cash from crypto fraud victims after digital transfers are blocked.

The vulnerability stems from a combination of deterministic bucket naming and insufficient ownership validation in the Vertex AI SDK. Attackers could preemptively create buckets in their own projects, intercepting victim model uploads. By replacing legitimate models with malicious ones during a narrow window, they achieve RCE via pickle deserialization. This flaw highlights the risks of implicit trust in cloud SDKs and underscores the need for explicit ownership checks. Mitigations include updating to SDK v1.148.0, using custom staging buckets, and monitoring for unauthorized bucket creations. The attack path demonstrates how subtle design flaws can escalate to full infrastructure compromise in multi-tenant environments.

Palo Alto Unit 42 · 2026-06-16 · Vulnerability: Attackers could hijack Vertex AI model uploads via predictable bucket names, leading to RCE in victim environments.

The CRA's SBOM requirement marks a pivotal shift from voluntary best practices to regulatory compliance, forcing organizations to scrutinize their software supply chains. Attack paths exploiting opaque dependencies—such as compromised third-party libraries—will become harder to conceal, but incomplete SBOMs create blind spots. Organizations must prioritize automated SBOM generation integrated into CI/CD pipelines, coupled with vendor SBOM procurement clauses in contracts. The focus should shift from mere compliance to actionable SBOM utilization, such as real-time vulnerability matching against NVD feeds. Mitigation requires cross-functional collaboration between legal, procurement, and DevSecOps teams to enforce SBOM completeness and accuracy.

Help Net Security · 2026-06-16 · Policy: SBOM adoption is accelerating due to CRA mandates, but supplier transparency and data quality remain significant hurdles.

The attack path involves threat actors uploading malicious wallpapers to Steam Workshop, which, when downloaded, execute scripts to harvest credentials or deploy malware. The scope includes any Steam user interacting with these wallpapers, potentially leading to account takeovers or further system compromise. Mitigations include enabling two-factor authentication, reviewing workshop subscriptions, and using endpoint protection to detect suspicious activity. Steam should enhance automated scanning of workshop content to flag or remove malicious uploads proactively.

Kaspersky Securelist · 2026-06-16 · Incidents: Malicious Steam Workshop wallpapers compromise gamer accounts through deceptive scripts.

The vulnerability in Cisco Catalyst SD-WAN Manager stems from inadequate input validation during file uploads, enabling authenticated attackers to manipulate filesystem contents via crafted HTTP requests. This flaw is particularly concerning as it can be chained with credential theft or insider threats to achieve root access. The attack path involves uploading malicious WAR files to specific directories, which are then deployed and executed. Affected deployments span on-prem, cloud, and government environments, amplifying the risk landscape. Mitigation requires immediate patching to the fixed versions listed in Cisco's advisory. Organizations should also audit logs for the provided IOCs, restrict write access to critical directories, and monitor for unusual file upload activities. The involvement of APT actor UAT-8616 in similar exploits underscores the need for heightened vigilance.

The Hacker News · 2026-06-16 · Vulnerability: Cisco patches an actively exploited SD-WAN Manager flaw (CVE-2026-20262) allowing file overwrites, with root escalation potential.

The vulnerability stems from improper symlink handling in the LiteSpeed cPanel plugin, allowing attackers with low-privilege access to escalate to root on shared hosting environments. This is particularly dangerous for CloudLinux/CageFS users, as it bypasses isolation mechanisms. The attack path likely involves chaining symlink manipulation with API calls (e.g., generateEcCert and packageUserSize) to trigger privilege escalation. Detection requires analyzing logs for specific patterns, with concurrent API calls serving as a key indicator. Mitigation involves upgrading to patched versions and auditing logs for suspicious activity. The lack of public exploit details suggests targeted attacks, emphasizing the need for proactive defense.

The Hacker News · 2026-06-16 · Vulnerability: CISA warns of actively exploited LiteSpeed cPanel plugin flaw (CVE-2026-54420) enabling root privilege escalation on shared hosting servers.

The UK's proposed ban on social media for children under 16 represents a significant shift in online safety policy, targeting platforms with user-generated content and algorithmic feeds. The inclusion of gaming sites and livestreaming restrictions expands the scope beyond typical social media platforms. Age assurance mechanisms will be critical, requiring robust verification methods to prevent circumvention. Privacy risks arise from potential data collection during age verification. Mitigation strategies should include transparent age-check processes, minimal data retention, and clear enforcement protocols. The political landscape may delay or dilute the ban, emphasizing the need for interim measures like parental controls and digital literacy programs.

The Record by Recorded Future · 2026-06-16 · Policy: UK to enforce the world's strictest online safety measures by banning social media for under-16s and restricting AI chatbots for under-18s.