A dark side to LLMs. [Research Saturday]
CyberWire Daily8 Apr 2023

A dark side to LLMs. [Research Saturday]

Sahar Abdelnabi from CISPA Helmholtz Center for Information Security sits down with Dave to discuss their work on "A Comprehensive Analysis of Novel Prompt Injection Threats to Application-Integrated Large Language Models." There is currently a large advance in the capabilities of Large Language Models or LLMs, as well as being integrated into many systems, including integrated development environments (IDEs) and search engines. The research states, "The functionalities of current LLMs can be modulated via natural language prompts, while their exact internal functionality remains implicit and unassessable." This could lead them to be susceptible to targeted adversarial prompting, as well as making them adaptable to even unseen tasks. Researchers demonstrated these said attacks to see if the LLMs needed new techniques for more defense. The research can be found here: More than you've asked for: A Comprehensive Analysis of Novel Prompt Injection Threats to Application-Integrated Large Language Models

Det här avsnittet är hämtat från ett öppet RSS-flöde och publiceras inte av Podme. Det kan innehålla reklam.

Avsnitt(3704)

How sovereign is Europe's space industry? [T-Minus: Space-Cyber Briefing]

How sovereign is Europe's space industry? [T-Minus: Space-Cyber Briefing]

As space becomes an increasingly important part of global communications, national security, and critical infrastructure, European governments are confronting a difficulty: How much control do they ne...

26 Juli 0s

Cold lures, hot targets. [Research Saturday]

Cold lures, hot targets. [Research Saturday]

This week, we are joined by Ondrej Kubovič, Security Awareness Specialist from ESET, discussing their work on "FrostyNeighbor: Fresh mischief and digital shenanigans." Ondrej walks us through ESET's l...

25 Juli 0s

Laundry Bear gets the spin cycle.

Laundry Bear gets the spin cycle.

Laundry Bear snuffles through unpatched Zimbra Collaboration servers. The State Department puts visa restrictions on cybercriminals. Oracle drops a record 1,449 security patches. Researchers disclose ...

24 Juli 0s

Do not pass Go(ogle).

Do not pass Go(ogle).

Google gets a billion dollar fine from the EU. The White House considers sanctions against Chinese AI developers. The GAO criticizes overlap in cyber reporting regulations. The Feds warn of Iranian ag...

23 Juli 0s

The AI has entered the chat.

The AI has entered the chat.

GPT escapes the sandbox and hacks Huggingface. SolarWinds patches multiple critical flaws. CISA orders patching of a critical Langflow AI vulnerability. A Paidwork breach affects over 23 million users...

22 Juli 0s

The defense against the AI arts.

The defense against the AI arts.

Trump's latest AI leader resigns. The Army burns through its AI tokens. Scammers impersonate IC3 personnel.HollowGraph malware uses a compromised Microsoft 365 calendar for C2. Qilin ransomware target...

21 Juli 0s

Behind the friendly face.

Behind the friendly face.

Hugging Face reports an autonomous AI-powered breach. Ernst & Young discloses a client data breach. Attackers are actively exploiting a critical ServiceNow flaw. Ransomware gangs sharpen their tactics...

20 Juli 0s

Europe's push for space sovereignty. [T-Minus: Space-Cyber Briefing]

Europe's push for space sovereignty. [T-Minus: Space-Cyber Briefing]

As space becomes an increasingly critical part of modern infrastructure, governments are reevaluating decades of policy to ensure reliable, secure, and independent access to the systems they are incre...

19 Juli 0s

Populärt inom Politik & nyheter

svenska-fall
p3-krim
aftonbladet-daily
aftonbladet-krim
flashback-forever
rss-sanning-konsekvens
tv4-nyheterna-story
rss-krimstad
rss-krimreportrarna
motiv
de-fyras-gang
rss-frandfors-horna
rss-vad-fan-hande
mannen-utan-spar
politiken
rss-flodet
spar
rss-aftonbladet-krim
krimmagasinet
grans