Threats Tagged 'generative ai'

A suspected Iran-nexus threat actor, dubbed Dust Specter, targeted Iraqi government officials in January 2026. The campaign involved impersonating Iraq's Ministry of Foreign Affairs and using compromised government infrastructure to host malicious payloads. Two attack chains were identified, utilizing previously undocumented malware including SPLITDROP, TWINTASK, TWINTALK, and GHOSTFORM. The malware employed creative evasion techniques, leveraged generative AI for development, and used file-based polling mechanisms for command execution. The campaign also incorporated ClickFix-style attacks and social engineering lures. Attribution to an Iran-nexus group is based on code similarities, victimology, and overlapping tactics with known Iranian APT groups.

MediumMalwareIraqIranUnited States+5#generative ai#twintask#twintalk

ESET researchers have discovered PromptSpy, the first known Android malware to abuse generative AI in its execution flow. This malware uses Google's Gemini AI to analyze screen content and provide instructions for UI manipulation, allowing it to adapt to various devices and layouts. PromptSpy's main purpose is to deploy a VNC module for remote access to the victim's device. It also abuses the Accessibility Service to block uninstallation, captures lockscreen data, and records video. The campaign appears to target users in Argentina and was likely developed in a Chinese-speaking environment. PromptSpy demonstrates how incorporating AI tools can make malware more dynamic and capable of real-time decision-making, potentially expanding the pool of potential victims.

MediumMalwareArgentinaChinaUnited States+3#remote access#gemini#generative ai

A new campaign targeting Indian government entities was uncovered, utilizing three backdoors: SHEETCREEP, FIREPOWER, and MAILCREEP. These tools leverage legitimate cloud services like Google Sheets, Firebase, and Microsoft Graph API for command and control, enabling the attackers to blend in with normal traffic. The campaign, named Sheet Attack, employed PDFs and malicious LNK files as initial infection vectors. Evidence suggests the use of generative AI in malware development. While sharing similarities with APT36, the campaign's unique characteristics point to either a new Pakistan-linked group or an APT36 subgroup. The attackers demonstrated hands-on-keyboard activity and deployed additional payloads, including a document stealer, to selected targets.

MediumMalwareIndiaUnited KingdomGermany+3#cloud services#sheetcreep#government

This report describes how generative AI techniques were utilized to accelerate the reverse engineering of XLoader malware, specifically version 8.0. By combining cloud-based static analysis of IDA exported data with dynamic checks, researchers rapidly unpacked encrypted code, deobfuscated API calls, and decrypted strings and domain names. The analysis uncovered three distinct function encryption schemes and a complex domain generation algorithm used by XLoader. The AI-assisted approach significantly reduced analysis time from days to hours, enabling faster extraction of indicators of compromise (IoCs). Despite AI's assistance, human expertise remained essential for overcoming the most sophisticated protections. The report highlights that generative AI can serve as a force multiplier for malware analysis, though malware authors may adapt their techniques in response. The threat is assessed as medium severity, with no known exploits in the wild currently. Several IoCs including hashes and suspicious domains are provided for detection and blocking.

MediumMalwareGermanyFranceNetherlands+4#generative ai#reverse engineering#chatgpt

Threat actors are leveraging generative AI tools like DeepSite AI and BlackBox AI to create phishing templates that closely mimic official Brazilian government websites, such as the State Department of Traffic and Ministry of Education. These malicious replicas are boosted in search results using SEO poisoning techniques. The phishing pages collect sensitive personal data, including CPF numbers and addresses, validating the information through APIs to build credibility. The ultimate goal is to trick victims into making payments via Pix, Brazil's instant payment system. Technical analysis reveals AI-generated source code signatures, including TailwindCSS styling, explanatory comments, and non-functional elements. The campaign demonstrates the evolving sophistication of phishing attacks empowered by generative AI tools.

MediumCampaignPortugalSpainItaly+3#deepsite ai#blackbox ai#pix payment