HiddenLayer warns that Google's Artificial Intelligence (AI) assistant Gemini faces indirect prompt injection flaws that could lead to phishing and chatbot takeover attacks. Indirect injections involve delivering the prompt injection via channels such as documents, emails, and other assets accessed by the Large Language Model (LLM), with the goal of taking over the model.

Following the discovery that thousands of US Congress staffers could be vulnerable to account hijacking and phishing, security experts have repeatedly warned against using work email addresses to sign up for third-party sites. The secure mail provider Proton collaborated with Constella Intelligence to search the dark web for over 16,000 publicly available email addresses associated with congressional staff. About 3,191 employees' emails were leaked to the dark web following third-party data breaches, with 1,848 listed together with plaintext passwords.

According to the Cybersecurity and Infrastructure Security Agency (CISA), a critical authentication bypass vulnerability in Ivanti Virtual Traffic Manager (vTM) has now been exploited by threat actors in the wild.  CISA added the bug to its long list of Known Exploited Vulnerabilities (KEV) on September 24, with federal agencies given until October 15 to patch it.  However, Ivanti has yet to update its security advisory to reflect the new information.

The US House Homeland Security Committee Republicans have recently unveiled a new bill aimed at addressing the growing cyber threats posed by state-sponsored Chinese actors targeting US critical infrastructure.  The legislation established an interagency task force led by the Cybersecurity and Infrastructure Security Agency (CISA) and the FBI.  The task force will focus on countering malicious cyber activity from the Chinese Communist Party (CCP), including advanced persistent threats (APTs) like Volt Typhoon.

A security researcher recently launched a project with the goal of showing that physical access control vulnerabilities still impact many organizations. The researcher noted that he documented nearly 40 instances of buildings that last year had hackable door controllers. He is now going through all the findings again to determine which of the buildings are still vulnerable, considering that more than a year has passed.

HP intercepted an email campaign involving a standard malware payload delivered via an Artificial Intelligence (AI)-generated dropper. The application of Generative AI (GenAI) on the dropper is a significant step towards novel AI-generated malware payloads. The company found an invoice-themed phishing email with an encrypted HTML attachment in June 2024. Phishers typically send targets a ready-encrypted archive file. However, in this case, the attacker implemented the AES decryption key in JavaScript in the attachment, which is uncommon.

The cyber espionage malware called "RomCom," which targeted the Ukraine military and its supporters last year, has returned with a new variant. It uses valid code-signing certificates to evade detection. Attackers can execute commands and download more malicious files in a multi-stage attack. The variant, dubbed "SnipBot" by researchers at Palo Alto's Unit 42, has been spreading since December. The malware, based on RomCom 3.0., also shares techniques seen in RomCom 4.0, thus making it the fifth version of the original RomCom Remote Access Trojan (RAT) family.

Mobile banking users worldwide are at risk from "Octo2," a new, advanced "Octo" malware variant. ThreatFabric analysts say Octo malware is one of the most widespread mobile threats in recent years. Octo2's advanced remote access and evasion features make it harder for security systems to detect. Its main enhancements involve increasing remote access stability, a critical feature in device takeover attacks. This article continues to discuss findings regarding the Octo2 malware variant.

An infostealer malware operation spanning 30 campaigns targeting different demographics and system platforms has been attributed to "Marko Polo," a cybercriminal group. The threat actors spread 50 malware payloads, including "AMOS," "Stealc," and "Rhadamanthys," via malvertising, spearphishing, and brand impersonation. According to Recorded Future's Insikt Group, the malware campaign has affected thousands, potentially resulting in millions of dollars in losses. This article continues to discuss new findings regarding the Marko Polo malware operation.

The cybersecurity community started warning about remote hacker attacks on Automatic Tank Gauge (ATG) systems nearly a decade ago, but critical vulnerabilities remain. ATG systems are widely deployed in gas stations, monitoring the parameters in a storage tank, including volume, pressure, and temperature. In 2015, some cybersecurity companies found that ATGs could be remotely hacked, warning that honeypot data revealed hackers had targeted these devices. Earlier this year, researchers at Bitsight found that the situation regarding vulnerabilities and exposed devices has not improved.