AI-Assisted Hacker Breaches 600 FortiGate Firewalls in 5 Weeks

Low-skilled cybercriminals are leveraging AI to enhance their extortion campaigns, a technique dubbed 'vibe extortion' by researchers. This involves using large language models (LLMs) to script professional extortion strategies, including deadlines and pressure tactics. While the attackers themselves lack technical depth, AI provides coherence and professionalism to their threats. The use of AI in cybercrime has evolved beyond simple grammar improvements to include rapid vulnerability scanning, parallelized targeting, and automated ransomware tasks. Unit 42 researchers highlight that AI acts as a 'force multiplier' for attackers, significantly reducing operational friction and lowering the barrier to entry for cybercriminals.

Adversarial AI-based attacks, such as those by Scattered Spider, are accelerating and leveraging living-off-the-land methods to spread and evade detection. These attacks use AI orchestration to perform network reconnaissance, discover vulnerabilities, move laterally, and harvest data at speeds that overwhelm manual detection methods. The Cloud Security Alliance report highlights over 70 ways autonomous AI-based agents can attack enterprise systems, expanding the attack surface beyond traditional security practices. Network Detection and Response (NDR) systems are increasingly being adopted to counter these AI-driven threats by providing real-time monitoring, analyzing network data, and identifying abnormal traffic patterns. NDR solutions can detect fast-moving, polymorphic attacks, summarize network activities, and render verdicts on potential threats, reducing the pressure on SOC analysts. Recent reports from Google's Threat Intelligence Group and Anthropic have revealed new AI-fueled attack methods, including the use of LLMs to generate malicious scripts and AI-orchestrated cyber espionage campaigns. Adversaries are also exploiting AV exclusion rules and using steganography techniques to evade detection. The combined use of NDR and EDR is essential for detecting and mitigating these sophisticated attacks.

Google's Threat Intelligence Group (GTIG) has identified new malware families that leverage artificial intelligence (AI) and large language models (LLMs) for dynamic self-modification during execution. These malware families, including PromptFlux, PromptSteal, FruitShell, QuietVault, and PromptLock, demonstrate advanced capabilities for evading detection and maintaining persistence. PromptFlux, an experimental VBScript dropper, uses Google's LLM Gemini to generate obfuscated VBScript variants and evade antivirus software. It attempts persistence via Startup folder entries and spreads laterally on removable drives and mapped network shares. The malware is under development or testing phase and is assessed to be financially motivated. PromptSteal is a data miner written in Python that queries the LLM Qwen2.5-Coder-32B-Instruct to generate one-line Windows commands to collect information and documents in specific folders and send the data to a command-and-control (C2) server. It is used by the Russian state-sponsored actor APT28 in attacks targeting Ukraine. State-backed hackers from China (APT31, Temp.HEX), Iran (APT42), North Korea (UNC2970), and Russia have used Gemini AI for all stages of an attack, including reconnaissance, phishing lure creation, C2 development, and data exfiltration. Chinese threat actors used Gemini to automate vulnerability analysis and provide targeted testing plans against specific US-based targets. Iranian adversary APT42 leveraged Gemini for social engineering campaigns and to speed up the creation of tailored malicious tools. The use of AI in malware enables adversaries to create more versatile and adaptive threats, posing significant challenges for cybersecurity defenses. Various threat actors, including those from China, Iran, and North Korea, have been observed abusing AI models like Gemini across different stages of the attack lifecycle. The underground market for AI-powered cybercrime tools is also growing, with offerings ranging from deepfake generation to malware development and vulnerability exploitation.

Microsoft's Digital Defense Report 2025 highlights a dramatic escalation in AI-driven cyber attacks. Microsoft systems analyze over 100 trillion security signals daily, indicating the growing sophistication and volume of cyber threats. Adversaries are leveraging generative AI to automate phishing, scale social engineering, and discover vulnerabilities faster than humans can patch them. Autonomous malware adapts tactics in real-time to bypass security systems, and AI tools themselves are becoming high-value targets. Microsoft's AI-powered defenses have reduced response times from hours to seconds, but defenders must remain vigilant as AI increases the speed and impact of cyber operations. Identity compromise remains a dominant attack vector, with phishing and social engineering accounting for 28% of breaches. Multi-factor authentication (MFA) prevents over 99% of unauthorized access attempts, but adoption rates are uneven. The rise of infostealers has fueled credential-based intrusions. The United States accounted for 24.8% of all observed attacks between January and June 2025, followed by the United Kingdom, Israel, and Germany. Government agencies, IT providers, and research institutions were among the most frequently targeted sectors. Ransomware remains a primary threat, with over 40% of recent cases involving hybrid cloud components.

The BRICKSTORM malware, attributed to PRC state-sponsored actors, has been used for long-term espionage against U.S. organizations, particularly in the technology, legal, SaaS, and BPO sectors. The malware, a Go-based backdoor, has been active for over a year, with an average dwell time of 393 days. It targets appliances without EDR support, such as VMware vCenter/ESXi, and uses legitimate traffic to mask its C2 communications. The attackers aim to exfiltrate emails and maintain stealth through various tactics, including removing the malware post-operation to hinder forensic investigations. The malware serves multiple functions, including web server, file manipulation, dropper, SOCKS relay, and shell command execution. The attackers use a malicious Java Servlet Filter (BRICKSTEAL) on vCenter to capture credentials and clone Windows Server VMs to extract secrets. The stolen credentials are used for lateral movement and persistence, including enabling SSH on ESXi and modifying startup scripts. The malware exfiltrates emails via Microsoft Entra ID Enterprise Apps, utilizing its SOCKS proxy to tunnel into internal systems and code repositories. UNC5221 focuses on developers, administrators, and individuals tied to China's economic and security interests. Mandiant has released a free scanner script to help defenders detect BRICKSTORM. CISA, NSA, and Cyber Centre issued a joint report on BRICKSTORM, providing IOCs, detection signatures, and recommended mitigations. The report highlights BRICKSTORM's advanced functionality to conceal communications, move laterally, and tunnel into victim networks. The malware automatically reinstalls or restarts if disrupted, and PRC actors are primarily targeting government and IT sector organizations. CISA analyzed eight BRICKSTORM samples from victim organizations and urges organizations to contact CISA if they detect BRICKSTORM or related activity. CISA warns that Chinese hackers have been backdooring VMware vSphere servers with Brickstorm malware, using multiple layers of encryption and a self-monitoring function to maintain persistence. The attackers compromised a web server in an organization's DMZ in April 2024, moved laterally to an internal VMware vCenter server, and deployed malware. They also hacked two domain controllers and exported cryptographic keys after compromising an ADFS server, maintaining access from at least April 2024 through September 2025. The attackers captured Active Directory database information and performed system backups to steal legitimate credentials and other sensitive data. CrowdStrike linked these attacks to a Chinese hacking group it tracks as Warp Panda, which also deployed previously unknown Junction and GuestConduit malware implants in VMware ESXi environments. The U.S. Cybersecurity and Infrastructure Security Agency (CISA) on Thursday released details of a backdoor named BRICKSTORM that has been put to use by state-sponsored threat actors from the People's Republic of China (PRC) to maintain long-term persistence on compromised systems. BRICKSTORM is a sophisticated backdoor for VMware vSphere and Windows environments, enabling cyber threat actors to maintain stealthy access and providing capabilities for initiation, persistence, and secure command-and-control. Written in Golang, the custom implant gives bad actors interactive shell access on the system and allows them to browse, upload, download, create, delete, and manipulate files. The malware, mainly used in attacks targeting governments and information technology (IT) sectors, also supports multiple protocols, such as HTTPS, WebSockets, and nested Transport Layer Security (TLS), for command-and-control (C2), DNS-over-HTTPS (DoH) to conceal communications and blend in with normal traffic, and can act as a SOCKS proxy to facilitate lateral movement. The cybersecurity agency did not disclose how many government agencies have been impacted or what type of data was stolen. The activity represents an ongoing tactical evolution of Chinese hacking groups, which have continued to strike edge network devices to breach networks and cloud infrastructures. In a statement shared with Reuters, a spokesperson for the Chinese embassy in Washington rejected the accusations, stating the Chinese government does not "encourage, support or connive at cyber attacks.". BRICKSTORM was first documented by Google Mandiant in 2024 in attacks linked to the zero-day exploitation of Ivanti Connect Secure zero-day vulnerabilities (CVE-2023-46805 and CVE-2024-21887). The use of the malware has been attributed to two clusters tracked as UNC5221 and a new China-nexus adversary tracked by CrowdStrike as Warp Panda. Earlier this September, Mandiant and Google Threat Intelligence Group (GTIG) said they observed legal services, software-as-a-service (SaaS) providers, Business Process Outsourcers (BPOs), and technology sectors in the U.S. being targeted by UNC5221 and other closely related threat activity clusters to deliver the malware. A key feature of the malware, per CISA, is its ability to automatically reinstall or restart itself by means of a self-monitoring function that allows its continued operation in the face of any potential disruption. In one case detected in April 2024, the threat actors are said to have accessed a web server inside an organization's demilitarized zone (DMZ) using a web shell, before moving laterally to an internal VMware vCenter server and implanting BRICKSTORM. However, many details remain unknown, including the initial access vector used in the attack and when the web shell was deployed. The attackers have also been found to leverage the access to obtain service account credentials and laterally move to a domain controller in the DMZ using Remote Desktop Protocol (RDP) so as to capture Active Directory information. Over the course of the intrusion, the threat actors managed to get the credentials for a managed service provider (MSP) account, which was then used to jump from the internal domain controller to the VMware vCenter server. CISA said the actors also moved laterally from the web server using Server Message Block (SMB) to two jump servers and an Active Directory Federation Services (ADFS) server, exfiltrating cryptographic keys from the latter. The access to vCenter ultimately enabled the adversary to deploy BRICKSTORM after elevating their privileges. CrowdStrike, in its analysis of Warp Panda, said it has detected multiple intrusions targeting VMware vCenter environments at U.S.-based legal, technology, and manufacturing entities this year that have led to the deployment of BRICKSTORM. The group is believed to have been active since at least 2022. Warp Panda exhibits a high level of technical sophistication, advanced operations security (OPSEC) skills, and extensive knowledge of cloud and virtual machine (VM) environments. Warp Panda demonstrates a high level of stealth and almost certainly focuses on maintaining persistent, long-term, covert access to compromised networks. Evidence shows the hacking group gained initial access to one entity in late 2023. Also deployed in the attacks alongside BRICKSTORM are two previously undocumented Golang implants, namely Junction and GuestConduit, on ESXi hosts and guest VMs, respectively. Junction acts as an HTTP server to listen for incoming requests and supports a wide range of capabilities to execute commands, proxy network traffic, and interact with guest VMs through VM sockets (VSOCK). GuestConduit, on the other hand, is a network traffic-tunneling implant that resides within a guest VM and establishes a VSOCK listener on port 5555. Its primary responsibility is to facilitate communication between guest VMs and hypervisors. Initial access methods involve the exploitation of internet-facing edge devices to pivot to vCenter environments, either using valid credentials or abusing vCenter vulnerabilities. Lateral movement is achieved by using SSH and the privileged vCenter management account "vpxuser." The hacking crew has also used the Secure File Transfer Protocol (SFTP) to move data between hosts. Some of the exploited vulnerabilities are listed below - CVE-2024-21887 (Ivanti Connect Secure), CVE-2023-46805 (Ivanti Connect Secure), CVE-2024-38812 (VMware vCenter), CVE-2023-34048 (VMware vCenter), CVE-2021-22005 (VMware vCenter), CVE-2023-46747 (F5 BIG-IP). The entire modus operandi revolves around maintaining stealth by clearing logs, timestomping files, and creating rogue VMs that are shut down after use. BRICKSTORM, masquerading as benign vCenter processes, is employed to tunnel traffic through vCenter servers, ESXi hosts, and guest VMs. Similar to details shared by CISA, CrowdStrike noted that the attackers used their access to vCenter servers to clone domain controller VMs, possibly in a bid to harvest the Active Directory Domain Services database. The threat actors have also been spotted accessing the email accounts of employees who work in areas that align with Chinese government interests. Warp Panda likely used their access to one of the compromised networks to engage in rudimentary reconnaissance against an Asia Pacific government entity. They also connected to various cybersecurity blogs and a Mandarin-language GitHub repository. Another significant aspect of Warp Panda's activities is their focus on establishing persistence in cloud environments and accessing sensitive data. Characterizing it as a "cloud-conscious adversary," CrowdStrike said the attackers exploited their access to entities' Microsoft Azure environments to access data stored in OneDrive, SharePoint, and Exchange. In at least one incident, the hackers managed to get hold of user session tokens, likely by exfiltrating user browser files and tunneled traffic through BRICKSTORM implants to access Microsoft 365 services via a session replay attack and download SharePoint files related to the organization's network engineering and incident response teams. The attackers have also engaged in additional ways to set up persistence, such as by registering a new multi-factor authentication (MFA) device through an Authenticator app code after initially logging into a user account. In another intrusion, the Microsoft Graph API was used to enumerate service principals, applications, users, directory roles, and emails. The adversary primarily targets entities in North America and consistently maintains persistent, covert access to compromised networks, likely to support intelligence-collection efforts aligned with PRC strategic interests. Chinese-speaking threat actors are suspected to have leveraged a compromised SonicWall VPN appliance as an initial access vector to deploy a VMware ESXi exploit that may have been developed as far back as February 2024. The exploit targeted three VMware vulnerabilities: CVE-2025-22224, CVE-2025-22225, and CVE-2025-22226. The toolkit includes simplified Chinese strings, suggesting a Chinese-speaking developer. The exploit uses Host-Guest File System (HGFS) for information leaking and Virtual Machine Communication Interface (VMCI) for memory corruption. The toolkit involves multiple components, including 'exploit.exe' (MAESTRO), 'devcon.exe', and 'MyDriver.sys', which write three payloads into VMX's memory: Stage 1 shellcode, Stage 2 shellcode, and VSOCKpuppet. VSOCKpuppet is a 64-bit ELF backdoor that provides persistent remote access to the ESXi host. The threat actors use 'client.exe' (GetShell Plugin) to send commands to the compromised ESXi host. The GetShell Plugin supports file transfer and command execution features. The toolkit prioritizes stealth over persistence.