Private Information Retrieval and Oblivious RAM

Private Information Retrieval and Oblivious RAM

🎙 Elaine Shi 👥 75K 📅 June 25, 2025 ⏱ 74 min 👁 734 📄 expert opinion 🧭 2026-08-06
Available in: English (current) Français

Keywords

ORAMPIRaccess patternprivacysecure computation

Summary

Elaine Shi presents a survey of Oblivious RAM (ORAM) and Private Information Retrieval (PIR), focusing on their practical applications and recent advances. She begins by motivating the need for access pattern hiding with real-world examples: Signal’s private contact discovery, blockchain private order flow to prevent front-running, and privacy-preserving large language models using retrieval-augmented generation. She explains that simply encrypting data is insufficient because access patterns leak sensitive information. She distinguishes ORAM from PIR: ORAM is suitable for outsourcing scenarios with per-client state, while PIR is better for global public databases. She notes that industry currently relies on trusted execution environments (TEEs) for efficiency, but the goal is to achieve similar performance without trusted hardware using cryptographic techniques like 2PC or PIR. She highlights that ORAM implementations, such as those from her company Oblivious Labs, can achieve microsecond-level query times on realistic database sizes (100GB to 1TB). The talk also covers theoretical progress, including client-preprocessing PIR, and discusses challenges such as handling database updates. Overall, the talk provides a comprehensive overview of the state of the art and future directions.

179 words

Critical Evaluation

The talk provides a valuable and up-to-date survey of ORAM and PIR, emphasizing practical deployment and recent theoretical advances. Elaine Shi is a leading expert, and her presentation is clear and well-structured. The real-world applications (Signal, blockchain, LLMs) effectively illustrate the importance of access pattern hiding. The distinction between ORAM and PIR is well explained, and the discussion of TEEs as a current practical solution, while acknowledging their limitations, is balanced. The talk does not delve into full technical details or proofs, but it is intended as a survey for a knowledgeable audience. The speaker mentions her own company’s implementation, which could be seen as a potential conflict of interest, but she also cites other work and discusses open challenges. The content is scientifically sound, and the sources mentioned (e.g., the empirical paper on Ethereum centralization) are credible. The talk does not include a formal comparison of all existing schemes, but it provides a good overview. The adéquation between title and content is excellent. Overall, this is a high-quality talk that would be useful for researchers and practitioners interested in privacy-enhancing technologies.

182 words

Title / Content Match

The title accurately reflects the content, which covers both Private Information Retrieval and Oblivious RAM, their differences, applications, and recent progress.

Quality & Reliability

8/10

Talk by a leading researcher in the field, presenting a survey of ORAM and PIR with real-world applications and recent results. The content is technically accurate and well-structured, but as a survey talk, it does not provide full proofs or experimental details. The speaker is a recognized expert, and the talk is hosted by a reputable institution (Simons Institute).

Key Moments

Cited Sources

Concurring Sources

  • Simons Institute talk page — The talk is hosted by the Simons Institute, a reputable institution, and the speaker is a leading researcher.

Contribution & Novelties

This talk provides a comprehensive and current overview of ORAM and PIR, bridging theory and practice. It highlights recent deployments (Signal, Meta) and ongoing efforts in blockchain and AI, and discusses the trade-offs between TEE-based solutions and purely cryptographic approaches. The speaker also presents recent performance results from her own company, indicating that ORAM can be practical at scale.

Pour aller plus loin :

101 words

Radar Profile

The radar profile shows high scores in quantity and quality of information, with slightly lower technical depth, reflecting a survey talk that is informative but not highly technical. The reliability score is high, consistent with the speaker's expertise and the institutional setting.

Reliability 8/10