MT-JailBench: A Modular Benchmark for Understanding Multi-Turn Jailbreak Attacks

Multi-turn jailbreaks exploit the ability of large language models to accumulate and act on conversational context. Instead of stating a harmful request directly, an attacker can gradually steer the conversation toward an unsafe answer. Recent methods demonstrate this risk, but they are usually...

AdapTools: Adaptive Tool-Based Indirect Prompt Injection Attacks on Agentic LLMs

The integration of external data services e.g., Model Context Protocol, MCP has made large language model-based agents increasingly powerful for complex task execution. However, this advancement introduces critical security vulnerabilities, particularly indirect prompt injection IPI attacks...

One Leak Away: How Pretrained Model Exposure Amplifies Jailbreak Risks in Finetuned LLMs

Finetuning pretrained large language models LLMs has become the standard paradigm for developing downstream applications. However, its security implications remain unclear, particularly regarding whether finetuned LLMs inherit jailbreak vulnerabilities from their pretrained sources. We investigat...

Enhanced MLLM Black-Box Jailbreaking Attacks and Defenses

Multimodal large language models MLLMs comprise of both visual and textual modalities to process vision language tasks. However, MLLMs are vulnerable to security-related issues, such as jailbreak attacks that alter the model's input to induce unauthorized or harmful responses. The incorporation o...

RAG Safety: Exploring Knowledge Poisoning Attacks to Retrieval-Augmented Generation

Retrieval-Augmented Generation RAG enhances large language models LLMs by retrieving external data to mitigate hallucinations and outdated knowledge issues. Benefiting from the strong ability in facilitating diverse data sources and supporting faithful reasoning, knowledge graphs KGs have been...

PYSEC-2023-214

Home assistant is an open source home automation. The audit team’s analyses confirmed that the redirecturi and clientid are alterable when logging in. Consequently, the code parameter utilized to fetch the accesstoken post-authentication will be sent to the URL specified in the aforementioned...

Design/Logic Flaw

Home assistant is an open source home automation. The audit team’s analyses confirmed that the redirecturi and clientid are alterable when logging in. Consequently, the code parameter utilized to fetch the accesstoken post-authentication will be sent to the URL specified in the aforementioned...

CVE-2023-41893 Account takeover via auth_callback login in Home Assistant Core

Home assistant is an open source home automation. The audit team’s analyses confirmed that the redirecturi and clientid are alterable when logging in. Consequently, the code parameter utilized to fetch the accesstoken post-authentication will be sent to the URL specified in the aforementioned...

The patch is not sufficient: there is another insidious exploit that can cause the same critical consequences

Lines of code Vulnerability details Status Has been reported to and confirmed by Jeff ENS team Note to the Judge I am not sure whether I should label this as a newly-identified High or a mitigation hard error. The root cause of this issue seems as same as the original report, but this requires us...

The Law of Unintended Outbreak – Who Is at Risk from Petya?

Hot on the heels of the global WannaCry outbreak in May, yesterday saw a wave of what looked like copycat malware sweeping the globe again. However, on closer inspection there may more to this than meets the eye, more than a simple new variant of an already established ransomware borrowing...

Sofacy APT organization to develop new Flash exploit framework-vulnerability warning-the black bar safety net

Sofacy cyber espionage Group, also known as Fancy Bear, and APT28, a Sednit, a Pawn Storm, and Strontium in. The organization has developed a new hack tool, and in the summer of this year the attacks have been put into use. Palo Alto Networks, the company said, the gang will be targeted in the...

POP3 Server type and version

The remote POP3 servers leak information about the software it is running, through the login banner. This may assist an attacker in choosing an attack strategy. Versions and types should be omitted where possible. OpenVAS Vulnerability Test $Id: popserverdetect.nasl 8023 2017-12-07 08:36:26Z teis...

Clearswift MIMEsweeper Manager Console Detection

The remote host appears to be running MIMEsweeper for SMTP, connections are allowed to the web MIMEsweeper manager console. Letting attackers know that you are using this software will help them to focus their attack or will make them change their strategy. C Tenable Network Security, Inc...