AI-assisted scientific knowledge presentation · experimental

Scientific Knowledge Maps

A paper should be readable at several depths: a research question and central answer in seconds, a hierarchy of claims and assumptions in minutes, and a source-linked evidence trail when verification matters.

For people

Hierarchical maps replace a single linear reading path with collapsible branches for claims, methods, evidence, limitations, artifacts, and scrutiny.

For agents

The same record is available as JSON with stable identifiers, typed relations, source anchors, provenance, and explicit epistemic status.

For honest uncertainty

Six dimensions remain separate. Missing evidence is visible, assumptions qualify claims, and no single score is presented as truth or correctness.

First mapped paper

VRASED: A Verified Hardware/Software Co-Design for Remote Attestation

VRASED combines verified HMAC software with a small hardware monitor, then connects component properties to end-to-end soundness and security under explicit MCU, compiler, cryptographic, and physical-attack assumptions.

Stage: author approved
Curation: approved by the author
27 stable map nodes
17 source anchors

Explore the human map · Read the JSON record

Complete Phase 1 catalog

All 79 paper entries

Every publication now has a stable map and six-axis profile. Summary- and abstract-bounded maps expose their incomplete source-audit status directly; VRASED is the author-reviewed, source-linked exemplar with page-level evidence and post-publication scrutiny.

Methods & Foundations

Secure Computation, Secret Sharing & Encrypted Computation 25

AI & Machine Learning 3

#79
Can Composing Generative Models Avoid Hallucinations? Implications for Cybersecurity Use Cases 2026 · Under review
Abstract-grounded map
#30
Automated Inference of Dependencies of Network Services and Applications via Transfer Entropy 2016 · Published algorithm
Abstract-grounded map
#16
Harvesting SSL Certificate Data to Identify Web-Fraud 2012 · Published · journal article algorithm
Source-linked map

Algorithms & Foundations 11

#67
Boosting the Performance of High-Assurance Cryptography: Parallel Execution and Optimizing Memory Access in Formally Verified Line-Point Zero-Knowledge 2023 · Published algorithm
Source-linked map
#63
How Byzantine Is a Send Corruption? 2022 · Published
Source-linked map
#56
Optimizing Registration Based Encryption 2021 · Published scheme
Source-linked map
#60
Quantum Optimization Heuristics with an Application to Knapsack Problems 2021 · Published algorithm
Source-linked map
#35
Efficient, Reusable Fuzzy Extractors from LWE 2017 · Published primitive · scheme
Source-linked map
#22
Neighborhood Watch: On Network Coding Throughput and Key Sharing 2013 · Published algorithm
Source-linked map
#11
Link-Layer Encryption Effect on Achievable Capacity in Wireless Network Coding 2010 · Published · workshop paper
Source-linked map
#8
Filtering Sources of Unwanted Traffic Based on Blacklists 2008 · Published · workshop paper algorithm
Source-linked map
#6
Optimal Filtering for DDoS Attacks 2007 · Published · workshop paper algorithm
Source-linked map
#1
Sub-Carrier Allocation Using Channel Prediction for OFDMA Systems Based on IEEE 802.16 Standard 2006 · Published algorithm
Source-linked map
#77
Towards Further Realizing Random Oracles: Post-Quantum Non-Malleable Point Obfuscation · Ongoing work
AI map draft

Security Domains

Privacy & Identity 17

Secure Systems & Networks 23

The six dimensions

A profile, not a verdict

Epistemic evidence: How directly and rigorously the work's important claims are supported, including the scope and limits of that support.
Auditability: Whether readers can inspect the paper's evidence, artifacts, assumptions, and validation path.
Production provenance: How clearly the human, AI, tool, revision, and approval process that produced the work is documented.
External scrutiny: The breadth and independence of review, reproduction, criticism, correction, or adversarial examination.
Reception: Dated evidence of independent use, extension, discussion, replication, teaching, or research lineage.
Contribution significance: The degree to which the work adds a first-of-kind capability, resolves an important problem, changes a belief, or opens a research direction.

Auditability follows the paper resources represented on this site: High when a publicly inspectable full-text copy is linked through an archive or author-hosted route, Medium when only official publication metadata is linked, and Low when no paper-specific resource is represented. A High rating means the paper can be inspected; it does not by itself establish artifact completeness, exact version correspondence, or independent reproduction.

Production provenance defaults to Medium when named authorship and publication or review status establish a baseline lifecycle trail. That default remains provisional when contributor roles, revision and effort history, AI or tool use, artifact-version lineage, or explicit final approval have not been audited.

Reception uses a dated, reproducible citation snapshot: Low for 0–8 located citations, Medium for 9–10, and High for 11 or more. Each paper records the count, source or search method, date, and coverage limits. This is a time-dependent attention signal, not evidence that the paper is correct.

Contribution significance captures first-of-kind capabilities, breakthroughs, and important resolutions without conflating them with correctness, reception, or scrutiny.

Contribution types are a separate controlled vocabulary: protocol for interactive procedures, primitive for foundational security functionality, scheme for algorithm/interface families, and algorithm for non-interactive computational procedures. Multiple types may apply.

Machine-readable index of all 79 paper entries · Machine-readable index of deep knowledge maps

Karim Eldefrawy

Scientific curiosity

Explore the connections

Scientists

Ideas

Karim’s work