MAI-2025-0007

Multimodal Large Language Models MLLMs are susceptible to attacks based on Jailbreak Probability, known as Jailbreak-Probability-based Attacks JPA. These attacks utilize a specialized Jailbreak Probability Prediction Network JPPN to detect and refine adversarial perturbations within input images...

New U.S. DoJ Rule Halts Bulk Data Transfers to Adversarial Nations to Protect Privacy

The U.S. Department of Justice DoJ has issued a final rule carrying out Executive Order EO 14117, which prevents mass transfer of citizens' personal data to countries of concern such as China including Hong Kong and Macau, Cuba, Iran, North Korea, Russia, and Venezuela. "This final rule is a...

Why Cybercriminals Are Not Necessarily Embracing AI

As published in HackerNoon and featured as a “Top 20 Best Read Article” for AI. Introduction The rapid advancement of AI has offered powerful tools for malware detection, but it has also introduced new avenues for adversarial attacks. As an example, recently OpenAI reported threat actors abusing...

MAI-2024-0017

Large Language Models LLMs utilizing Reinforcement Learning from Human Feedback RLHF for safety alignment are susceptible to a sophisticated "alignment-based" jailbreak attack. This attack employs a best-of-N sampling strategy in conjunction with an adversarial LLM to efficiently craft prompts th...

AIs Discovering Vulnerabilities

I've been writing about the possibility of AIs automatically discovering code vulnerabilities since at least 2018. This is an ongoing area of research: AIs doing source code scanning, AIs finding zero-days in the wild, and everything in between. The AIs aren't very good at it yet, but they're...

A Sherlock Holmes Approach to Cybersecurity: Eliminate the Impossible with Exposure Validation

Sherlock Holmes is famous for his incredible ability to sort through mounds of information; he removes the irrelevant and exposes the hidden truth. His philosophy is plain yet brilliant: "When you have eliminated the impossible, whatever remains, however improbable, must be the truth." Rather tha...

Researchers Reveal 'Deceptive Delight' Method to Jailbreak AI Models

Cybersecurity researchers have shed light on a new adversarial technique that could be used to jailbreak large language models LLMs during the course of an interactive conversation by sneaking in an undesirable instruction between benign ones. The approach has been codenamed Deceptive Delight by...

The Growing Importance of Exposure Management: Our Key Insights from Gartner® Hype Cycle™ for Security Operations, 2024

The Gartner® Hype Cycle™ for Security Operations, 2024 was published in late July, and is an interesting look at the dynamic nature of both the threat landscape and the diverse range of technologies that security & risk management SRM professionals use to safeguard their organizations...

Exploit for Out-of-bounds Write in Facebook Netconsd

CVE-2023-28753 An integer overflow vulnerability tha...

CTEM in the Spotlight: How Gartner's New Categories Help to Manage Exposures

Want to know what's the latest and greatest in SecOps for 2024? Gartner's recently released Hype Cycle for Security Operations report takes important steps to organize and mature the domain of Continuous Threat Exposure Management, aka CTEM. Three categories within this domain are included in thi...

Cybersecurity Agencies Warn of China-linked APT40's Rapid Exploit Adaptation

Cybersecurity agencies from Australia, Canada, Germany, Japan, New Zealand, South Korea, the U.K., and the U.S. have released a joint advisory about a China-linked cyber espionage group called APT40, warning about its ability to co-opt exploits for newly disclosed security flaws within hours or...

CVE-2024-6382 Adversarial unsanitized input may cause MongoDB Rust Driver to issue unintended commands.

Incorrect handling of certain string inputs may result in MongoDB Rust driver constructing unintended server commands. This may cause unexpected application behavior including data modification. This issue affects MongoDB Rust Driver 2.0 versions prior to 2.8.2...

CVE-2024-6382 Adversarial unsanitized input may cause MongoDB Rust Driver to issue unintended commands.

Incorrect handling of certain string inputs may result in MongoDB Rust driver constructing unintended server commands. This may cause unexpected application behavior including data modification. This issue affects MongoDB Rust Driver 2.0 versions prior to 2.8.2...

Adversarial unsanitized input may cause MongoDB Rust Driver to issue unintended commands.

Incorrect handling of certain string inputs may result in MongoDB Rust driver constructing unintended server commands. This may cause unexpected application behavior including data modification. This issue affects MongoDB Rust Driver 2.0 versions prior to 2.8.2...

Prompt Injection Flaw in Vanna AI Exposes Databases to RCE Attacks

Cybersecurity researchers have disclosed a high-severity security flaw in the Vanna.AI library that could be exploited to achieve remote code execution vulnerability via prompt injection techniques. The vulnerability, tracked as CVE-2024-5565 CVSS score: 8.1, relates to a case of prompt injection...

Chinese Cyber Espionage Targets Telecom Operators in Asia Since 2021

Cyber espionage groups associated with China have been linked to a long-running campaign that has infiltrated several telecom operators located in a single Asian country at least since 2021. "The attackers placed backdoors on the networks of targeted companies and also attempted to steal...

adversarial-robustness-toolbox (1.6.1), aiqc (>=2.1.0, <=7.1.5) +120 more potentially affected by CVE-2024-5206 via scikit-learn (0.24.1)

scikit-learn PYPI version 0.24.1 is affected by a known vulnerability. The following packages have a transitive dependency on scikit-learn and may be impacted: - adversarial-robustness-toolbox 1.6.1 - aiqc =2.1.0, =0.7.0, =2.0.0, =0.0.37, =0.0.1, =0.1.0, =2021.3.29.16, =0.2.4, =0.1.0, =0.0.396,...

adversarial-robustness-toolbox (1.6.1), aiqc (>=2.1.0, <=7.1.5) +120 more potentially affected by CVE-2024-5206 via scikit-learn (0.24.1)

scikit-learn PYPI version 0.24.1 is affected by a known vulnerability. The following packages have a transitive dependency on scikit-learn and may be impacted: - adversarial-robustness-toolbox 1.6.1 - aiqc =2.1.0, =0.7.0, =2.0.0, =0.0.37, =0.0.1, =0.1.0, =2021.3.29.16, =0.2.4, =0.1.0, =0.0.396,...

MAI-2024-0055

Large Vision Language Models LVLMs are susceptible to a sophisticated attack known as the bi-modal adversarial prompt BAP. This attack exploits the integration of textual and visual prompts to circumvent established safety protocols, thereby eliciting harmful outputs from models engineered to...

New Lawsuit Attempting to Make Adversarial Interoperability Legal

Lots of complicated details here: too many for me to summarize well. It involves an obscure Section 230 provision--and an even more obscure typo. Read this...