12 stories · 3 sources · 2 critical · 4 high · ~13 min read
Coverage: Last 24 hours
Today’s Highlights
Active exploitation in cryptocurrency and webmail platforms highlights the week, as Bitget suffered a $351 million loss linked to a backend vulnerability and Roundcube’s CVE-2026-48842 SQL injection remains a favored attacker target. Major SaaS and IaaS providers like Salesforce and Cloudflare face scrutiny for CRM exfiltration and container data leakage, respectively. Defenders must also address pervasive side-channel leaks in file notification APIs and adapt policy for AI misuse as both a technical and legal challenge.
Table of Contents
- Bitget Says Suspected North Korean Hackers Stole $351.6M After Backend Compromise
- Roundcube Pre-Auth SQL Injection Flaw Actively Exploited in the Wild
- ‘SalesBleed’ Flaws in Salesforce Agentforce Enabled Zero-Click Data Exfiltration
- Cloudflare Fixes Flaw That Let One Container Read Another Customer’s Leftover Disk Data
- Unpatched OnePlus Flaws Let Installed Android Apps Gain Root Without Permissions
- Windows, Linux, Android File Notification Systems Leak User Activity
- WSO2 and Adobe Commerce Flaws Exploited in Attacks, Added to CISA KEV
Critical High Medium Low
Top Stories
Bitget Says Suspected North Korean Hackers Stole $351.6M After Backend Compromise
Source: The Hacker News | Published: Sep 25 | Risk: CRITICAL | Impacted: Cryptocurrency exchanges, Fintech companies with digital wallets, DeFi platforms | Topics: Vulnerability / Exploit
What happened: Bitget reported that suspected North Korean hackers stole $351.6 million from its hot and warm wallets on September 24, 2026. The attack exploited a backend system vulnerability to authorize unauthorized fund transfers. Bitget has suspended withdrawals and is conducting a comprehensive security review with third-party assistance.
Why it matters: Backend compromise of cryptocurrency custody systems can result in massive financial losses and regulatory risk, especially when attackers bypass normal transaction controls.
How it works: Cryptocurrency exchanges maintain wallets with varying exposure (hot, warm, or cold), and backend systems authenticate and authorize fund transfers. A backend flaw allowed attackers to trigger unauthorized movements from wallets, bypassing user-facing controls.
Practitioner Perspective
Crypto exchanges must assume that backend vulnerabilities represent high-consequence, low-dwell-time risks: attackers only need a brief window of access to irreversibly move funds from hot or warm wallets. Attribution to North Korean actors aligns with historical campaigns targeting digital asset platforms for rapid theft and laundering. Given the scale, incident response runbooks and forensic checkpoints should be prepared before such events occur. The top priority for defenders is to implement withdrawal limits, strong segregation of duties, and continuous behavioral monitoring on wallet transaction flows, not just classic perimeter defenses.
Recommended Actions
- Implement transaction-level anomaly detection in wallet management backends, focusing on irregular patterns and large or unusual withdrawals
- Review and test withdrawal freeze and kill-switch procedures to contain future backend breaches
Roundcube Pre-Auth SQL Injection Flaw Actively Exploited in the Wild
Source: The Hacker News | Published: Sep 25 | Risk: HIGH | Impacted: Self-hosted Roundcube webmail operators, Universities and NGOs using open-source email, Small and mid-market businesses with on-prem mail | Topics: Vulnerability / Exploit
What happened: A pre-authentication SQL injection vulnerability in Roundcube Webmail’s virtuser_query plugin (CVE-2026-48842) is being actively exploited, allowing unauthenticated attackers to inject arbitrary SQL statements, potentially exposing mail account credentials and stored messages. Patches were released in May 2026.
Why it matters: Active exploitation of unauthenticated SQL injection in mail platforms directly exposes account credentials and sensitive message data to remote attackers.
How it works: Roundcube Webmail is a PHP-based IMAP mail client widely used for web access to email. A vulnerability in the virtuser_query plugin allows unauthenticated SQL injection, meaning remote attackers can manipulate the database before any user authentication.
Affected / Fix: Patched in May 2026 for all supported Roundcube versions using virtuser_query.
Practitioner Perspective
Any public-facing Roundcube Webmail deployment running a vulnerable virtuser_query plugin is currently a high-value target. Attackers can compromise mailboxes at scale, use them for internal phishing or pivot to other systems, and leverage privileged access gained from extracted credentials. Security teams with legacy or unpatched mail deployments will face difficult incident response if attackers have already accessed logs or message stores. Defenders must prioritize patching and conduct forensic review for signs of exploitation, not just clean-and-patch.
Recommended Actions
- Apply patches from May 2026 for Roundcube CVE-2026-48842 to all instances using the virtuser_query plugin
- Hunt for suspicious SQL queries and mailbox access patterns in Roundcube and database logs since exploitation was reported
‘SalesBleed’ Flaws in Salesforce Agentforce Enabled Zero-Click Data Exfiltration
Source: SecurityWeek | Published: Sep 25 | Risk: HIGH | Impacted: Salesforce customers using Agentforce, Businesses with integrated Web-to-Lead workflows, Customer support centers leveraging CRM automations | Topics: Vulnerability / Cloud
What happened: Researchers identified three vulnerabilities in Salesforce’s Agentforce platform, collectively termed ‘SalesBleed,’ which could allow attackers to hijack trusted agents, exfiltrate sensitive CRM data, and launch phishing attacks. These flaws were exploitable via Web-to-Lead forms, enabling zero-click data exfiltration and phishing through Agentforce agents.
Why it matters: Flaws in CRM platform integrations can result in silent, large-scale data exfiltration, exposing sensitive business and customer information without user involvement.
How it works: Salesforce Agentforce manages digital agents and workflows within the Salesforce ecosystem. Weaknesses in input validation and agent-trusting logic allowed attackers to hijack agents and export sensitive customer relationship management (CRM) data via legitimate API calls.
Practitioner Perspective
Organizations leveraging Salesforce Agentforce need to understand that Web-to-Lead forms and agent integrations may have enabled attackers to exfiltrate data with no user action. These channels can be abused to stage attacks inside trusted business workflows, including phishing and automatic data theft. Zero-click vectors are particularly risky, because security teams may lack visibility or alerting on data moving through legitimate cloud channels. The key task is to audit integrated agents and remediate trust relationships that enabled or could enable similar exposure.
Recommended Actions
- Assess exposure to SalesBleed flaws in Salesforce Agentforce platforms, specifically reviewing Web-to-Lead and agent configuration
- Audit agent permissions and restrict unnecessary read/export access to sensitive CRM fields
Cloudflare Fixes Flaw That Let One Container Read Another Customer’s Leftover Disk Data
Source: The Hacker News | Published: Sep 25 | Risk: HIGH | Impacted: Public cloud customers on container-backed infrastructure, SaaS providers hosting regulated data, Organizations with strict data destruction requirements | Topics: Vulnerability / Cloud
What happened: A flaw in Cloudflare Containers let a paying customer read data that other customers’ containers had left behind on the same server, Cloudflare and the researchers who found it said on Thursday. The data came from disk space that earlier containers had used and given up, not from any live workload, and an attacker could not choose whose data they
Why it matters: Data residue across container tenants weakens isolation, creating regulatory and confidentiality risk for organizations with sensitive workloads in multi-tenant public clouds.
How it works: Containers are lightweight, isolated application environments often sharing physical hosts in cloud platforms. The bug permitted a customer to read disk blocks previously used by another (now-removed) tenant, exposing data not intended for sharing.
Affected / Fix: Cloudflare states the flaw is fixed; no details on affected versions in source.
Practitioner Perspective
Defenders leveraging container-based SaaS or IaaS need to be aware that storage layer bugs can defeat hard-won logical isolation. If attackers can recover disk data from previous tenants, even ‘not-live’ leftovers may include sensitive or regulated information. This risk is magnified for industries subject to retention, privacy, or sector-specific standards (e.g., financial, healthcare, legal). Proactive review of your supply chain and cloud workload isolation claims is critical, particularly for forensic or audit-sensitive organizations. Demand explicit statements of isolation and sanitization in your vendor assurances.
Recommended Actions
- Validate that Cloudflare workloads and similar SaaS vendors employ secure container teardown and disk scrubbing procedures
- Request or perform forensic testing for disk data persistence post-container teardown in your production environments
Unpatched OnePlus Flaws Let Installed Android Apps Gain Root Without Permissions
Source: The Hacker News | Published: Sep 24 | Risk: HIGH | Impacted: OnePlus mobile device users, Enterprise fleets with OnePlus phones, BYOD environments | Topics: Ics Ot / Exploit
What happened: A OnePlus 15 running the latest OxygenOS can be rooted by a malicious app the owner installs, one that asks for no special permissions. A researcher, Rasmus Moorats, chained two flaws in OnePlus’s own software to gain root access, the highest level of control over an Android phone. OnePlus told him the same flaws affect many more of its own devices and
Why it matters: Unpatched privilege escalation vulnerabilities in widely-deployed mobile devices enable malicious apps to silently obtain root access, eroding trust in mobile fleet controls.
How it works: OnePlus devices run OxygenOS, a custom Android fork. Two chained vulnerabilities in device-specific software allow any app to escalate privileges to root without user or permission prompts.
Practitioner Perspective
Enterprise and consumer OnePlus users face elevated risk: any installed application, even one with minimal runtime permissions, may abuse these flaws to gain full device control. This bypasses nearly all mobile device management (MDM) and malware containment assumptions. Opportunistic attackers can persist, exfiltrate sensitive data, or stage lateral attacks on networks reachable from compromised devices. Immediate attention is needed for high-risk user groups with access to sensitive information via mobile platforms. Defensive focus should be on limiting further app installations and identifying indicators of unauthorized privilege escalation.
Recommended Actions
- Restrict installation of new or unvetted applications on affected OnePlus devices until patches are released
- Monitor device telemetry for signs of privilege escalation or unexpected root activity
Windows, Linux, Android File Notification Systems Leak User Activity
Source: SecurityWeek | Published: Sep 25 | Risk: MEDIUM | Impacted: Endpoint workstations, Shared desktop environments, Mobile devices, Research and high-privacy users | Topics: Vulnerability / Threat Intel
What happened: Researchers at Graz University of Technology demonstrated that file-change notification systems in Linux, Windows, Android, and macOS can be exploited to monitor user activities, including keystroke timing and browsing behavior, by observing file names and event timings without accessing file contents.
Why it matters: File notification APIs in common operating systems can leak sensitive details about user actions and application workflows, increasing exposure to monitoring and behavioral fingerprinting even without full compromise.
How it works: File change notification subsystems (such as inotify on Linux, fsevents on macOS, and FS events on Windows/Android) allow processes to watch for filesystem changes. Attackers can exploit these to infer what users are doing by monitoring event timings and filenames, even if file contents remain protected.
Practitioner Perspective
Anyone relying on local security boundaries or process isolation on Windows, Linux, Android, or macOS should revisit assumptions about privacy and telemetry controls. Attackers can pivot from a low-privilege foothold to infer user decisions, credentials entry, or sensitive operations simply by monitoring filesystem metadata events. This expands the attack surface for side-channel and user tracking campaigns, especially in high-target-value environments such as government or research. For organizations with privacy obligations or threat actors using stealthy surveillance tradecraft, these notification channels warrant explicit review. The most important task is to assess which endpoints or contexts could expose operational workflows via these filesystems and restrict unnecessary access accordingly.
Recommended Actions
- Audit use of inotify, fsevents, and similar file notification APIs for unintentional data exposure in sensitive Linux, macOS, and Android endpoints
- Restrict local access to file notification subsystems for untrusted users or processes
Exploits & CVEs
WSO2 and Adobe Commerce Flaws Exploited in Attacks, Added to CISA KEV
Source: The Hacker News | Published: Sep 25 | Risk: CRITICAL | Impacted: Enterprises running WSO2 API management platforms, Organizations with Adobe Commerce or Magento deployments, E-commerce operators | Topics: Vulnerability / Cloud
What happened: The U.S. Cybersecurity and Infrastructure Security Agency (CISA), on Thursday, added two critical security flaws impacting WSO2 and Adobe Commerce and Magento to its Known Exploited Vulnerabilities (KEV) catalog, based on evidence of active exploitation. The vulnerabilities are listed below – CVE-2026-5430 (CVSS 9.8) – A path traversal vulnerability in WSO2 API Control Plane,
Why it matters: Attackers are actively exploiting critical vulnerabilities in enterprise API and e-commerce platforms, directly threatening data integrity and system availability.
How it works: WSO2 is a widely used API gateway and control plane platform; Adobe Commerce (Magento) powers significant e-commerce operations. Path traversal vulnerabilities in these products let attackers manipulate file system access, often leading to arbitrary code execution or data theft.
Affected / Fix: Added to CISA KEV due to active exploitation; patch availability implied.
Practitioner Perspective
Organizations using WSO2 or Adobe Commerce/Magento frameworks should treat new exploitation as a clear signal to prioritize urgent patching and forensics. Path traversal and related bugs in API control planes or e-commerce frameworks often yield full system compromise, data theft, or facilitate subsequent supply chain attacks. Since CISA has added these issues to the KEV list, there is credible evidence of broad targeting or weaponization. The major risk is business interruption and customer data leakage: teams must prioritize verification of fixes and actively hunt for exploitation artifacts.
Recommended Actions
- Patch WSO2 installations for CVE-2026-5430 and Adobe Commerce/Magento for listed exploited CVEs immediately
- Search SIEM and application logs for indicators of path traversal and unauthorized requests targeting API endpoints
Also Today
- Roundcube Webmail Vulnerability in Attackers’ Crosshairs: CVE-2026-48842 remains a target in ongoing in-the-wild attacks on vulnerable Roundcube installations.
- ThreatsDay: AI Search Poisoning, AI Coding Tool Leaking Repos, One-Click Code Execution and 13 More Stories: Attackers exploit AI-driven search, code tools, and supply chain mechanisms to subvert user trust.
- Autonomous AI Hacks Raise Thorny Questions of Legal Accountability: Opaque legal responsibilities for AI-driven attacks create uncertainty for organizations and regulators.
- Kontext Security Emerges With $4 Million for AI Agent Runtime Controls: Startup launches new platform for runtime monitoring and control of autonomous AI agents.
- On Anthropic’s AI Misuse Report: Review details 117 recorded AI misuse cases including reconnaissance, propaganda, and dual-use risks.
Defensive Actions
- Apply May 2026 patches for Roundcube CVE-2026-48842 and conduct forensics on exposure
- Assess Salesforce Agentforce Web-to-Lead integrations and audit agent permissions
- Enforce strong segregation and detection controls in cryptocurrency wallet backends
- Review Cloudflare and similar IaaS environments for evidence of cross-tenant data persistence
- Restrict unvetted application installs on OnePlus devices; monitor for privilege escalation signs
- Audit file notification usage and restrict access on Linux, macOS, and Android systems
- Patch WSO2 for CVE-2026-5430 (CVSS 9.8) and Adobe Commerce/Magento deployments; hunt for indicators of compromise
- Update cloud and SaaS vendor due diligence to require explicit data isolation and disk scrubbing guarantees
What We’re Watching
- Additional exploitation of Roundcube CVE-2026-48842 and unpatched mail platforms
- Ongoing targeting of WSO2 CVE-2026-5430 (CVSS 9.8) and Adobe Commerce/Magento, especially with CISA KEV inclusion
- Patch release cycles and incident response for OnePlus OxygenOS privilege escalation
- New campaigns abusing Salesforce Agentforce or similar SaaS integrations for zero-click exfiltration
- Evidence of cross-tenant data persistence or side-channel telemetry leaks in mainstream public cloud container offerings
Found this briefing useful? Follow the blog to get the next one as soon as it is published, and pass it along to a colleague who owns patching.
Categories: Cybersecurity Blog, Cybersecurity News
Leave a Reply