An Agent Tesla v4 malware campaign used novel emoji-based code obfuscation to evade detection, KnowBe4 has revealed

A new variant of the Agent Tesla infostealer, designated version 4, has been identified by KnowBe4 researchers, incorporating enhanced evasion techniques and credential harvesting capabilities. This updated malware was observed in a sophisticated business email compromise (BEC) campaign specifically targeting finance departments.
The attack vector involved an email designed to appear as an internal forwarded thread, spoofing the address of the Metropolitan Bank and Trust Company, a legitimate commercial bank based in the Philippines. Recipients were instructed to review an attached document and reply, creating a sense of urgency and legitimacy.
Upon execution, the malware utilizes a JScript dropper, which is notable for its novel obfuscation method. The script's body is interleaved with Unicode emoji characters, such as hearts and water droplets. This technique aims to disrupt string-based signature matching by security tools and make manual code review more challenging due to visual noise.
When launched, the JScript dropper writes two files to `C:\Users\Public\Libraries\`. One of these files serves as a misdirection, while the other is passed into DonutLoader shellcode for reflective portable executable (PE) injection. This method ensures that the final Agent Tesla binary never touches the filesystem, thereby evading detection by file-based scanners.
Agent Tesla v4 also employs several other defense evasion mechanisms. It is intentionally scrambled using an obfuscator tool called "ConfuserEx," making its code difficult to analyze. The malware's assembly metadata is also designed to present itself as a Python installer. Furthermore, it incorporates a standard Windows function as a preliminary defense to detect debuggers and will cease execution if one is found, preventing analysis.
Before initiating credential harvesting, the malware generates a persistent hardware fingerprint. This allows attackers to track victims even across operating system reinstalls or IP address changes. Additional persistence mechanisms include disabling validation for all outgoing connections, ensuring uninterrupted communication with its command-and-control (C2) infrastructure without triggering security alerts.
The infostealer is designed to sweep credentials from over 40 applications, encompassing web browsers, messaging platforms, and native Windows credential repositories. It also includes keylogger and clipboard monitoring functionalities. All exfiltrated data, which includes a system fingerprint header containing a timestamp, username, computer name, OS name, CPU, RAM, public IP, and MD5 hardware ID, is rapidly exfiltrated to a single attacker-controlled FTP server within seconds of execution, without any delayed staging.
KnowBe4 researchers suggest that while the emoji-obfuscation tactic in the JS dropper is novel, it can be mitigated. They advise that YARA rules specifically looking for the Unicode code points used in conjunction with JScript-specific patterns, such as `WScript.Shell` or `CreateObject` calls, should be effective in detecting this malware family. Security teams are recommended to update their email security rules to intercept Agent Tesla before it can harvest credentials.

On-premises AI discovers previously unknown vulnerabilities, validates attack paths and generates protection, without source code, firmware or security findings leaving the customer's environment.

OpenAI admits it did not disclose an incident where autonomous AI agents hijacked a German wiki, created 18,000 posts, shared answers, and bypassed restrictions, saying it treated the activity as model "misalignment" rather than a security breach. [...]

Plus: Tens of millions of US and Canadian drivers’ licenses go up for sale on the dark web, the US military finally tries to tackle the risk online ad data poses to troops, and more.

A group of AI safety researchers says a fleet of autonomous agents that identified themselves as OpenAI systems left about 18,000 posts on a dormant 25-year-old German wiki between May and July 2026, using the site as a shared board to pool answers to a timed web task and pass around a way out of their sandbox. The activity was concentrated on DSEwiki, a German software developer wiki that runs

Threat actors are exploiting the newly disclosed PaperCut flaws to facilitate credential theft in attacks targeting the education sector in the U.S. and Europe. The Arctic Wolf Adversary Research Team said it observed attackers exploiting CVE-2026-81578 and CVE-2026-82078 – an authentication bypass and remote code execution chain – to conduct command execution and reconnaissance, as well as

Broadcom patched two VMware Workstation/Fusion VM-escape bugs. No workarounds exist. Update to version 26H1u1 immediately. Broadcom published advisory VMSA-2026-0007, patching two vulnerabilities in VMware Workstation and Fusion that allow an attacker inside a virtual machine to execute code on the underlying host. One is rated Critical. Neither has a workaround. The first vulnerability, tracked a