BEGIN:VCALENDAR
VERSION:2.0
PRODID:Data::ICal 0.24
BEGIN:VEVENT
DESCRIPTION:   'Title: Cove: Compositional and Verifiable Confidential Comp
 uting\n   Workflows\n   Tags: Crypto & Privacy Village | Creator Talk/Pane
 l\n   When: Sunday\, Aug 9\, 11:00 - 11:30 PDT\n   Where: LVCCW Level 1 Ha
 ll 3 1102 (Creator Stage 6) - [1]Map\n\n   Description:\n\n   Confidential
  computing systems involve computation over private inputs\n   held by mut
 ually distrusting parties while producing public\,\n   cryptographically v
 erifiable evidence of what was computed. Trusted\n   execution environment
 s (TEEs) are a practical building block for these\n   systems\, enabling c
 ode to run inside hardware-encrypted enclaves that\n   emit attestations b
 inding a measurement of the loaded code to genuine\n   hardware. We presen
 t Cove\, a framework that composes attested\n   computations into multi-st
 age\, multi-party workflows structured as\n   directed acyclic graphs. Eac
 h node runs in its own enclave\, decrypts\n   private inputs only under at
 testation-gated key release\, and emits a\n   certificate that downstream 
 nodes can require as a cryptographic\n   precondition before consuming ups
 tream artifacts\; anyone holding the\n   workflow bytes and the TEE vendor
 's attestation roots can verify the\n   whole chain non-interactively. We 
 show that a small set of reusable\n   primitives - confidential artifact p
 rovisioning\, encrypted dynamic\n   outputs\, attested ephemeral-key chann
 els (RA-TLS)\, and\n   certificate-gated preconditions - compose into syst
 ems that lift\n   attestation from verified code identity to verified runt
 ime\n   properties\, and use this to express practical applications in sec
 ure\n   AI systems including confidential AI inference\, attested AI\n   b
 enchmarks\, and training-code verification. We give an open-source\n   ref
 erence implementation on Intel TDX via Phala Cloud's dstack and\n   demons
 trate end-to-end feasibility with an attested benchmark\n   workflow.\n\n 
   Speakers:Stephanie\,Robin\,Erika Lee\n\n   SpeakerBio:  Stephanie\n\n   
 Stephanie is a research engineer working on AI security and safety.\n   He
 r research interests include adversarial robustness\, agent security\n   a
 nd verifiable AI deployments. Previously\, she was at Meta\, where she\n  
  worked on AI safety and security evaluations (CyberSecEval) and\n   guard
 rails (PromptGuard\, LlamaFirewall)\, and prior to that she worked\n   on 
 application security for web and mobile.\n\n   SpeakerBio:  Robin\n\n   Ro
 bin is a Member of Technical Staff at Inferact (maintainer of vLLM)\,\n   
 working towards the best inference stack of the future. Previously he\n   
 worked as a software engineer at Near Protocol\, Meta\, Google\, and\n   T
 witter. His interests include application security\, model\n   red-teaming
 \, distributed systems\, and GPU performance optimization.\n\n   SpeakerBi
 o:  Erika Lee\n\n   Erika Lee is a 3rd-year Math–Computer Science underg
 raduate student\n   at UC San Diego and a MATS 9.1 AI Security Fellow. Her
  research\n   interests are in verifiable and privacy-preserving AI\, and 
 building\n   secure and trustworthy AI systems. She previously interned wi
 th the US\n   Department of Defense's AI/ML portfolio and the U.S. Army\, 
 working on\n   deploying AI/ML systems in classified environments.\n\n   B
 ing-Jyue Chen is a 3rd-year PhD candidate in Computer Science at\n   UIUC.
  His current research focuses on zero knowledge machine learning\n   (zkml
 ). He is also interested in verifiable AI\, privacy-preserving\n   machine
  learning\, and advanced cryptography.\n\n   Daniel Kang is an assistant p
 rofessor at UIUC in the computer science\n   department. His research focu
 ses on making analytics with machine\n   learning easy for scientists and 
 analysts to use\, and has been\n   supported by Google\, the Open Philanth
 ropy project\, Emergent Ventures\,\n   and others. He has consulted at Ope
 nAI\, Google\, Microsoft\, hedge\n   funds\, and other companies to suppor
 t their ML deployments\, and\n   advised a number of startups.\n\n   '\n\n
    1. #LVCCW_Level1_Hall3\n\n\n
DTEND:20260809T183000Z
DTSTART:20260809T180000Z
LOCATION:Crypto & Privacy Village - LVCCW Level 1 Hall 3 1102 (Creator Stag
 e 6)
SUMMARY:Cove: Compositional and Verifiable Confidential Computing Workflows
END:VEVENT
END:VCALENDAR