@sahar_abdelnabi Profile picture

Sahar Abdelnabi 🍉🕊

@sahar_abdelnabi

She/her. AI Security Researcher at Microsoft (MSRC) | ex. PhD @CISPA | Neurodivergent 🧠🦋 | all things AI, safety, security | peace for all #CeasefireNOW

Similar User
@fraboeni photo

@fraboeni

Mario Fritz photo

@mariojfritz

Yang Zhang photo

@realyangzhang

ELSA - European Lighthouse on Secure and Safe AI photo

@elsa_lighthouse

Tobias Lorenz photo

@tobias_ml

Giancarlo Pellegrino photo

@tgianko

Aurore Fass photo

@AuroreFass

Vera Xinyue Shen photo

@xyshen365

Roberto Amoroso photo

@roberto__mrs

Rui WEN photo

@RuiWen_CISPA

Soheil photo

@Soheil__K

Sven Bugiel photo

@svebug

Junjie(Jony) Chu photo

@Junjie7Chu

Lea Schönherr photo

@leaschnherr

Kathrin Grosse photo

@KathrinGrosse

Pinned

Very excited about this work!! LLMs in applications process inputs from many sources, making them vulnerable to prompt injections. We look into models' internals (activations) to catch if models drifted from users' instructions after processing supposedly data-only sources. 1/

Tweet Image 1

Sahar Abdelnabi 🍉🕊 Reposted

If you're a security researcher, you might want to check out the Microsoft Zero Day Quest which launched today and runs through January 25, 2025. This is a public research challenge that expands our bug bounty programs and offers an additional $4 million in rewards for…


Sahar Abdelnabi 🍉🕊 Reposted

We are hiring! Multiple PhD and PostDoc positions in Trustworthy AI, AI Security, AI for Science, AI for Code, Causal Discovery, CySec, Privacy, ... cispa.saarland/group/fritz/ #ELLISPhD Application is a fantastic opportunity to apply (deadline Nov. 15th): bit.ly/3CqzCvi

Tweet Image 1

Sahar Abdelnabi 🍉🕊 Reposted

🧙 I am recruiting PhD students and postdocs to work together on making sure AI Systems and Agents are built safe and respect privacy (+ other social values). Apply to UMass Amherst @manningcics and enjoy a beautiful town in Western Massachusetts. Reach out if you have questions!

Tweet Image 1

Sahar Abdelnabi 🍉🕊 Reposted

I'm looking forward to hiring PhD students and postdocs to push the boundaries of AI security/safety/privacy @CSatETH If you're interested, please apply by **19 November 2024** to the ETHZ AI Center (link below). I'm also happy to chat with applicants at NeurIPS later this year


Sahar Abdelnabi 🍉🕊 Reposted

The IDF bombed a clearly marked vehicle and killed four engineers going to repair water infrastructure in southern Gaza after they had coordinated the trip with Israeli authorities, Oxfam says oxfam.org/en/press-relea…


When your paper goes through three rounds of reviews and rebuttals and you end up with 52-page camera-ready version 🙈🙀

Happy to share that our paper was (finally!!) accepted at #NeurIPS2024 D&B!! We propose dynamic, multi-agent, scorable negotiation games to assess ToM, planning, inference, arithmetic skills, and deception between agents. Our benchmark is highly extensible and evolving.

Tweet Image 1


Sahar Abdelnabi 🍉🕊 Reposted

We need a ceasefire for the elderly, the young, and the children. We need a ceasefire even for the animals and the trees. We need a ceasefire so Jews can celebrate their holidays in peace, and so non-Jews can have holidays in this region in peace too. We need a ceasefire to heal…


Sahar Abdelnabi 🍉🕊 Reposted

AISec 2024 update: We're delighted to share our complete workshop schedule, featuring 18 accepted papers, engaging poster sessions, and three distinguished speakers in AI security. Visit aisec.cc to explore the program!

Tweet Image 1

More details and the official launch soon!

📢Check out the @satml_conf competition "Adaptive Prompt Injection Challenge" The competition invites participants to evade prompt injection defences in an LLM-integrated email client. 😈Craft emails that execute tasks while avoiding detection. More at: microsoft.github.io/llmail-inject/



Sahar Abdelnabi 🍉🕊 Reposted

You should be so lucky to have people throughout your research career that you can openly bounce ideas to and from - especially if they complement your strengths in your areas of weakness - it is a rare and precious gift.

Doing good science is 90% finding a science buddy to constantly talk to about the project.

Tweet Image 1


October is ADHD awareness month

ADHD is NOT a "mild" condition. People dismiss and belittle the potential seriousness and severity so much. The effects of untreated, unmanaged, unrecognised ADHD can be life-threatening, and are (at least) hugely destructive.



Anyone had an issue before that an @arxiv paper doesn't appear on G scholar after changing the title? (New title gives nothing, old title points to non-arxiv results), any idea how to fix?


Sahar Abdelnabi 🍉🕊 Reposted

⭐️ NeurIPS D&B Spotlight ⭐️ We will be presenting the findings from our LLM CTF together with a multi-turn prompt extraction dataset at NeurIPS. Kudos to amazing colleagues @edoardo_debe @dpaleka @leaschnherr @sahar_abdelnabi @florian_tramer @mariojfritz

The report and dataset from our LLM CTF are finally out! Learn more about the winning strategies and the resulting dataset! 🧵 twitter.com/edoardo_debe/s…



Sahar Abdelnabi 🍉🕊 Reposted

Excited to share that our LM-GC has been accepted at #NeurIPS2024. Through lossless gradient compression, we show the potential of zero-shot LLMs as gradient priors: Accurate prior 👉 efficient compression. Stay tuned for future exciting applications. #DeepLearning #LLM

Tweet Image 1

POV: You are chilling after a deadline and minding your own business then @iclr_conf gives you a mini heartattack that your papers might be disk rejected because you didn't sign up for review while in fact you did 🙃


Sahar Abdelnabi 🍉🕊 Reposted

millions displaced. thousands dead. the numbers are numbing, but when you zoom in is when the heartbreak really hits. my grandfather has dementia. every day since my family has had to evacuate their home in Lebanon, he asks why he can’t go home 💔


Sahar Abdelnabi 🍉🕊 Reposted

🔍Exciting Opportunity in AI Security & Privacy! Our Azure Research team in Cambridge, UK, is hiring a Postdoc. Join us in developing mechanisms that provide robust guarantees against security & privacy threats to systems with unreliable AI components. jobs.careers.microsoft.com/global/en/job/…


Sahar Abdelnabi 🍉🕊 Reposted

My entire life, Beirut has been synonymous with war-torn. Arabs have been synonymous with terror. It shouldn’t matter but I want you to know that it’s a beautiful, complicated city filled with art and gorgeous food and ordinary people who take daily walks by the sea at sunset.

Tweet Image 1
Tweet Image 2

Loading...

Something went wrong.


Something went wrong.