SOBRE ESTE EPISÓDIO
Prompt injection demonstrates one of the major challenges in securing LLMs and agents -- how do you ensure an agent ignores attackers and only does what you instructed it to do. The flaw highlights how LLMs mix inputs, context, and outputs without any strict boundaries between them. Julie Brunias joins us to talk through examples of injections, why their consequences can go beyond information leaks, and why trying to mitigate them with other LLMs is insufficient.
Segment resources:
- https://llmgateway.io/open-source
- https://www.wiz.io/blog/off-guard-breaking-litellm-from-authentication-bypass-to-cloud-compromise
Show Notes: https://securityweekly.com/asw-401
Inglês
Estados Unidos da América
TRANSCRIÇÃO 🔗
Are you the producer of this podcast?
Add a podcast transcript
Need Audio-to-Text?
Transcribe with Listen411 in Just 60 Seconds
PESQUISAR EPISÓDIOS PASSADOS
Pesquisar episódios anteriores de Application Security Weekly (Video).
OUTROS EPISÓDIOS NESTE PODCAST
AppSec has always emphasized techniques and tools for discovering vulns, along with taxonomies and lists for describing them. But just piling up more CVEs into a prioritized patching queue has never been an effective strategy. Nidhi Aggarwal talks about some of the economics and decisions that orgs…
We showcase recordings from this year's Black Hat.
The Hidden Risks of the AI Supply Chain - Black Hat interview with Michael Leland, VP and Field CTO of Island
Agents can independently discover and install tools, but the emerging ecosystem of Skills and MCP servers lacks many of the trust and se…
All sorts of cybersecurity disciplines are adopting agents to help humans save time and automate routine activities. Sai Kiran Uppu describes his work on creating a platform for agents to analyze external threat intel, examine internal systems, and present triage decisions to operators. This type o…
While agents and LLMs haven't fundamentally changed core mobile vulnerability types, they have supercharged speed, scale, and accessibility—democratizing threats like automated phishing, synthetic identity fraud, and easier identification of hard-coded secrets. Ryan Lloyd and Jason Cortlund break d…
Sandboxing, least privilege, and monitoring are well-established controls in terms of the defenses they provide against unexpected and unauthorized actions. But being well-established in theory doesn't always translate to successful in practice. Kieran Human talks about some of the properties that …
Aviso: O podcast e a arte incorporada nesta página são de Mike Shema, que é propriedade de seu proprietário e não é afiliado ou endossado por Listen Notes, Inc.
EDITAR
Obrigado por ajudar a manter a base de dados dos podcasts atualizada.
