Google Adds Multi-Layered Defenses to Secure GenAI from Prompt Injection Attacks

Google has revealed the various safety measures that are being incorporated into its generative artificial intelligence AI systems to mitigate emerging attack vectors like indirect prompt injections and improve the overall security posture for agentic AI systems. "Unlike direct prompt injections,...

CVE-2019-8760

This issue was addressed by improving Face ID machine learning models. This issue is fixed in iOS 13. A 3D model constructed to look like the enrolled user may authenticate via Face ID...

Detecting Quishing Attacks with Machine Learning Techniques through QR Code Analysis

The rise of QR code based phishing "Quishing" poses a growing cybersecurity threat, as attackers increasingly exploit QR codes to bypass traditional phishing defenses. Existing detection methods predominantly focus on URL analysis, which requires the extraction of the QR code payload, and may...

Development of an Adapter for Analyzing and Protecting Machine Learning Models from Competitive Activity in the Networks Services

Due to the increasing number of tasks that are solved on remote servers, identifying and classifying traffic is an important task to reduce the load on the server. There are various methods for classifying traffic. This paper discusses machine learning models for solving this problem. However, su...

FLARE: Feature-Based Lightweight Aggregation for Robust Evaluation of IoT Intrusion Detection

The proliferation of Internet of Things IoT devices has expanded the attack surface, necessitating efficient intrusion detection systems IDSs for network protection. This paper presents FLARE, a feature-based lightweight aggregation for robust evaluation of IoT intrusion detection to address the...

AIs as Trusted Third Parties

This is a truly fascinating paper: "Trusted Machine Learning Models Unlock Private Inference for Problems Currently Infeasible with Cryptography." The basic idea is that AIs can act as trusted third parties: Abstract: We often interact with untrusted parties. Prioritization of privacy can limit t...

Malicious ML Models on Hugging Face Leverage Broken Pickle Format to Evade Detection

Cybersecurity researchers have uncovered two malicious machine learning ML models on Hugging Face that leveraged an unusual technique of "broken" pickle files to evade detection. "The pickle files extracted from the mentioned PyTorch archives revealed the malicious Python content at the beginning...

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...

The vulnerability of the library for optimizing machine learning models in Intel Neural Compressor allows attackers to exploit it by bypassing security measures related to SQL query structures, thereby enabling them to enhance their privileges.

The vulnerability of the Intel Neural Compressor library for optimizing machine learning models is related to the lack of protective measures for SQL query structures. Exploiting this vulnerability can allow attackers to enhance their privileges remotely...

Division By Zero Error

TensorFlow is vulnerable to a division by zero error. The vulnerability is due to insufficient handling of cases where the input's fourth dimension is zero in the DepthwiseConv TFLite operator, which can allows to execution issues or crashes in machine learning models...

Unpacking 2024's SaaS Threat Predictions

Early in 2024, Wing Security released its State of SaaS Security report, offering surprising insights into emerging threats and best practices in the SaaS domain. Now, halfway through the year, several SaaS threat predictions from the report have already proven accurate. Fortunately, SaaS Securit...

RunGptLLM class in LlamaIndex has a command injection

A command injection vulnerability exists in the RunGptLLM class of the llamaindex library, version 0.9.47, used by the RunGpt framework from JinaAI to connect to Language Learning Models LLMs. The vulnerability arises from the improper use of the eval function, allowing a malicious or compromised...

GHSA-PW38-XV9X-H8CH RunGptLLM class in LlamaIndex has a command injection

A command injection vulnerability exists in the RunGptLLM class of the llamaindex library, version 0.9.47, used by the RunGpt framework from JinaAI to connect to Language Learning Models LLMs. The vulnerability arises from the improper use of the eval function, allowing a malicious or compromised...

CVE-2024-4181

A command injection vulnerability exists in the RunGptLLM class of the llamaindex library, version 0.9.47, used by the RunGpt framework from JinaAI to connect to Language Learning Models LLMs. The vulnerability arises from the improper use of the eval function, allowing a malicious or compromised...

CVE-2024-4181 Command Injection in run-llama/llama_index

A command injection vulnerability exists in the RunGptLLM class of the llamaindex library, version 0.9.47, used by the RunGpt framework from JinaAI to connect to Language Learning Models LLMs. The vulnerability arises from the improper use of the eval function, allowing a malicious or compromised...

CVE-2024-4181

The CVE-2024-4181 issue affects llama_index (RunGptLLM class) version 0.9.47 used by RunGpt framework to connect to LLMs. Root cause: unsafe use of eval enabling a malicious or compromised LLM hosting provider to run arbitrary commands on a client machine. Impact statements in sources indicate th...

CVE-2024-4181 Command Injection in run-llama/llama_index

A command injection vulnerability exists in the RunGptLLM class of the llamaindex library, version 0.9.47, used by the RunGpt framework from JinaAI to connect to Language Learning Models LLMs. The vulnerability arises from the improper use of the eval function, allowing a malicious or compromised...

CVE-2024-34359

CVE-2024-34359 affects llama-cpp-python (Python bindings for llama.cpp). The vulnerability arises when init loads a model’s chat template from the gguf metadata and constructs self.chat_handler via llama_chat_format.Jinja2ChatFormatter.to_chat_handler(), using a sandbox-less Jinja2 Environment. R...

How to keep your ChatGPT conversations out of its training data

Last week, OpenAI announced it had given ChatGPT users the option to turn off their chat history. ChatGPT is a "generative AI", a machine learning algorithm that can understand language and generate written responses. Users can interact with it by asking questions, and the conversations users hav...

Google Cloud Introduces Security AI Workbench for Faster Threat Detection and Analysis

Google's cloud division is following in the footsteps of Microsoft with the launch of Security AI Workbench that leverages generative AI models to gain better visibility into the threat landscape. Powering the cybersecurity suite is Sec-PaLM, a specialized large language model LLM that's...