Oblivious RAM (ORAM) is a cryptographic technique that hides a program's memory access pattern—the sequence of read and write operations to physical storage—from an observer, such as an untrusted cloud server or a compromised operating system. By continuously reshuffling and re-encrypting data blocks, ORAM ensures that two identical logical access sequences produce indistinguishable physical access patterns, making it impossible for an adversary to infer sensitive information like search queries or branching logic from memory side-channels.
Glossary
Oblivious RAM (ORAM)
A cryptographic technique that hides a program's memory access pattern—which locations are read or written—from an observer, preventing data leakage through memory access side-channels.
MEMORY ACCESS PATTERN HIDING
What is Oblivious RAM (ORAM)?
A cryptographic primitive that conceals a program's memory access patterns from an untrusted server or observer, preventing data leakage through the observation of which physical memory locations are read or written.
Modern ORAM constructions, such as Path ORAM and Circuit ORAM, achieve this by organizing memory as a tree of encrypted buckets and reading an entire path on every access, then evicting and relocating the target block to a new random path. This imposes a polylogarithmic overhead on memory operations, making ORAM a critical building block for secure processors, private information retrieval, and multi-party computation systems where protecting access patterns is as vital as encrypting the data itself.
ARCHITECTURAL PRIMITIVES
Key Features of ORAM
Oblivious RAM transforms a program's memory access pattern into a sequence that is computationally indistinguishable from random, neutralizing a critical side-channel exploited by cache-timing and bus-snooping attacks.
01
Access Pattern Hiding
The foundational guarantee of ORAM is that an adversary observing physical memory addresses learns nothing about the program's true computation. Whether the program is reading a secret key or a public constant, the sequence of physical accesses is statistically or computationally indistinguishable from a random permutation. This prevents leakage through cache-timing attacks, cold-boot attacks, and bus snooping.
02
Recursive ORAM Construction
A critical optimization that eliminates the need for a large client-side position map. The position map itself is stored in a smaller ORAM, which is stored in an even smaller ORAM, recursively until the final map fits in O(1) client storage. This technique, pioneered by Shi et al., reduces client-side memory from O(N) to O(1) blocks while maintaining poly-logarithmic overhead.
03
Path ORAM
The simplest and most practical ORAM scheme to date. On each access, the client reads and rewrites an entire path from the root to a leaf in a binary tree stored on the server. The requested block is remapped to a new random leaf after each access. Path ORAM achieves O(log N) bandwidth overhead with small, constant client storage, making it the basis for most modern hardware and software implementations.
04
Tree-Based Framework
The dominant paradigm for constructing efficient ORAMs. Data blocks are stored in a binary tree on the server, with each block assigned to a random leaf path. A stash on the client temporarily holds overflowing blocks. Invariants ensure that a block is always found either on its assigned path or in the stash. This framework underpins Path ORAM, Circuit ORAM, and many variants.
05
Secure Processor Integration
ORAM is a critical primitive in trusted execution environments (TEEs) like Intel SGX and AMD SEV. While TEEs encrypt memory contents, they do not hide access patterns by default. Integrating a hardware-accelerated ORAM controller prevents an untrusted operating system or a physical bus snooper from inferring sensitive computations—such as cryptographic key schedules—from memory traces.
06
Bandwidth Overhead Trade-off
The central cost metric for any ORAM scheme. Trivial ORAM (scanning the entire memory on every access) achieves perfect security with O(N) overhead. Path ORAM reduces this to O(log² N) or O(log N) with larger blocks. Square-root ORAM offers a different trade-off with O(√N) overhead but simpler logic. The choice depends on the memory size and the acceptable latency budget.
ORAM FUNDAMENTALS
Frequently Asked Questions
Clear, technical answers to the most common questions about Oblivious RAM, its mechanisms, and its role in privacy-preserving computation.
Oblivious RAM (ORAM) is a cryptographic technique that completely hides a program's memory access pattern—which physical or logical addresses are read or written—from an observer who can see the memory bus. It works by continuously re-encrypting and physically shuffling data blocks on the storage medium after every access, ensuring that the sequence of observed memory locations is statistically independent of the actual computation. The seminal construction by Goldreich and Ostrovsky introduced a hierarchical architecture where data is stored in levels of increasing size, and a position map tracks the current location of each block. During a read or write, the client scans a small set of blocks, retrieves the target, and then re-inserts it at a new random location, making two consecutive accesses to the same logical address appear as accesses to completely unrelated physical addresses. Modern tree-based ORAM schemes, like Path ORAM, organize storage as a binary tree of buckets, reading and rewriting an entire path from the root to a leaf to hide which specific block was accessed.
Enabling Efficiency, Speed & Accuracy
Intelligent Analysis, Decision & Execution
We build AI systems for teams that need search across company data, workflow automation across tools, or AI features inside products and internal software.
Talk to Us
Search across company data
Give teams answers from docs, tickets, runbooks, and product data with sources and permissions.
Useful when people spend too long searching or get different answers from different systems.
Enterprise searchRAGPermissions
Read more
Automate internal workflows
Use AI to route work, draft outputs, trigger actions, and keep approvals and logs in place.
Useful when repetitive work moves across multiple tools and teams.
AI agentsWorkflow automationGovernance
Read more
Add AI to products and internal tools
Build assistants, guided actions, or decision support into the software your team or customers already use.
Useful when AI needs to be part of the product, not a separate tool.
AI integrationDecision supportModel routing
Read more
About the author
Prasad Kumkar
CEO & MD, Inference Systems
Prasad Kumkar is the CEO & MD of Inference Systems and writes about AI systems architecture, LLM infrastructure, model serving, evaluation, and production deployment. Over 5+ years, he has worked across computer vision models, L5 autonomous vehicle systems, and LLM research, with a focus on taking complex AI ideas into real-world engineering systems.
His work and writing cover AI systems, large language models, AI agents, multimodal systems, autonomous systems, inference optimization, RAG, evaluation, and production AI engineering.
LinkedIn
Limited slotsGet a Free AI Consultation
Partnered with leading AI, data, and software stack.
How We Work
Custom AI workflows for your Business
One-fit-all AI don't work for modern businesses. At Inferensys, we aim to understand your business & custom requirements; which we use to define most efficient agentic workflows, the data, and the tools for your business.
01
Review the use case
We understand the task, the users, and where AI can actually help.
Read more02
Pick the right approach
We define what needs search, automation, or product integration.
Read more03
Build the first useful version
We implement the part that proves the value first.
Read more04
Improve from there
We add the checks and visibility needed to keep it useful.
Read moreThe first call is a practical review of your use case and the right next step.
Talk to Us