A Scalable Trust Discovery Architecture for the Internet of Agents
Song Zhang, Jiankang Yao, Hongtao Li, Xiaojun Zhang, Xugang Shen, Xin Li, Xin Li, Yanbiao Li
cs.CR · cs.AI
Quick verdict
DNS-inspired Root-Registry-Resolver architecture for trusted agent discovery, with a fast prototype but no baselines and only qualitative security analysis.
Summary
This paper addresses a gap in the emerging Internet of Agents: existing protocols such as MCP and A2A handle communication between agents that already know each other, but scalable registration, verifiable identity, and capability-oriented discovery remain unsolved. Inspired by DNS, the authors propose a three-layer architecture — Agent Root (trust anchor governing trusted registry suffixes), Agent Registry (registration and metadata publication), and Agent Resolver (distributed, trust-aware discovery). A registry-suffix-anchored composite identity binds native agent identifiers (domains, DIDs, platform IDs) to trusted suffixes, e.g. abc.com.atiagent.cn. A dual-certificate scheme (private CA intra-domain, public CA cross-domain) supports three progressive authentication levels: mTLS, transparency-log checking, and DANE/DNSSEC binding. A prototype named ATI (Nginx front proxy plus root, registry, transparency-log and resolver services) was benchmarked with wrk on dual-CPU E5-2650 servers: 58ms mean registration latency (153ms p99), 25ms mean discovery latency (165ms p99), over 19,000 registration QPS and over 29,000 discovery QPS, with a case study on Alibaba's Tongyi Qianwen platform. Main limitations: no baseline comparison against AgentDNS, ANS, or centralized directories; a purely qualitative security analysis without threat model or attack experiments; underreported benchmark conditions (server counts, concurrency, workload mix); unspecified capability-matching semantics; and no code release.
Review complete
Assessment at a glance
4
risk flags identified
● high 0
● medium 3
● low 1
4 strengths · 5 weaknesses · 5 author questions
AI text signal = Binoculars zero-shot detector score over 20 sampled passages (Fast-DetectGPT co-reported) — absence does not prove human authorship, and a flag does not imply misconduct.
Levels reflect verification complexity — not a quality score. Evolve never rates your paper.
Key claims under review
- Current agent protocols (MCP, A2A) leave scalable registration, trustworthy identification, and capability-oriented discovery unresolved for the Internet of Agents.
- A Root-Registry-Resolver hierarchy with registry-suffix-anchored composite identities (F_a = Normalize(I_a).S_r) yields globally discoverable, DNS-compatible agent identities.
- A dual-certificate mechanism (private CA intra-domain, public CA cross-domain) with three progressive authentication levels (mTLS, transparency log, DANE/DNSSEC) strengthens inter-agent identity trust.
- The ATI prototype achieves 58ms mean registration and 25ms mean discovery latency and sustains over 19,000 registration QPS and over 29,000 discovery QPS, demonstrating feasibility and scalability.
Strengths
Well-motivated gap in a fast-moving area
Sections 1-2 precisely frame the registration/identity/discovery void left by MCP [22] and A2A [23] and situate the work against the right related work (AgentDNS [28], ANS [14,29]), which is the current competitive set.
DNS-compatible design reuses proven infrastructure
The composite identity (Sec. 3.2) stays inside the DNS hierarchy so publication, caching, and delegation piggyback on mature mechanisms; the Root maintains only a trusted-suffix catalog, a clean governance decomposition that avoids a global per-agent database.
Concrete, standards-based trust mechanism
End-to-end prototype with real-platform case study and honest scoping
Weaknesses
No baselines for the headline numbers
Benchmark methodology underreported
Security analysis is purely qualitative
Capability discovery semantics missing
Presentation sloppiness undermines precision
Full Review Report
$1.99 · one-timeUnlock the complete preliminary analysis Evolve produced for this paper.
Report preview
Scope & Claims
This is a systems/architecture paper from CNNIC and collaborators proposing ATI (agent trust infrastructure), a DNS-inspired three-layer design for registering, identifying, and discovering AI agents at Internet scale. The claimed gap is real and current: MCP [22] and A2A [23] presuppose that communicating agents already know each other, leaving registration, verifiable identity, and capability-oriented discovery unsolved (Sec. 1, Sec. 2.2). The four contributions (Sec. 1) are (1) a Root-Registry-Resolver hierarchy, (2) a registry-suffix-anchored composite identity F_a = Normalize(I_a).S_r (Sec. 3.2), (3) dual-certificate, three-level authentication (Sec. 3.3, Table 1), and (4) a prototype evaluated at 58ms/25ms mean latencies and >19k/>29k QPS (Sec. 4.3, Table 4). Claims are scoped to identity-level trust; Sec. 4.4 explicitly disclaims behavioral or service-quality guarantees — an unusually honest scoping statement.
The remaining sections cover Methodology Assessment, Statistical Evidence, Results & Reproducibility, Limitations & Risks, and Verdict & Recommendations.
- All 4 strengths & 5 weaknesses, analyzed in detail
- 4 risk flags with severity, section & reviewer questions
- Methods & statistical evidence assessment
- 5 questions a peer reviewer would ask
- Full six-dimension evaluation report & recommendations
One-time purchase · this paper only
One-time purchase, this paper only. AI-generated preliminary analysis — not peer review.
Frequently asked questions
What is the Internet of Agents and why does it need a trust discovery architecture?
The Internet of Agents is a proposed infrastructure layer where large numbers of autonomous AI agents, built by different organizations on different platforms, can find, verify, and collaborate with each other. This paper argues that current protocols like MCP and A2A only connect agents that already know each other, so a separate infrastructure for registration, verifiable identity, and capability search is needed — analogous to what DNS and PKI provide for the web.
How does the ATI architecture in this paper work?
ATI borrows DNS's hierarchy: an Agent Root maintains a catalog of trusted registry suffixes; Agent Registries (each owning one suffix) register agents and publish DNS records plus structured capability/trust metadata cards; Agent Resolvers synchronize that data and answer capability queries. Each agent receives a composite identity such as abc.com.atiagent.cn, verified through public/private CA certificates, transparency logs, and optionally DANE/DNSSEC binding.
What performance does the ATI prototype achieve?
In wrk benchmarks on servers with two Intel E5-2650 CPUs and 32GB RAM: agent registration averaged 58ms (127ms at the 95th percentile, 153ms at the 99th), agent discovery averaged 25ms (32ms p95, 165ms p99), with sustained throughput above 19,000 registration QPS and 29,000 discovery QPS. DNS record publication averaged 9ms and naming review 4ms. These are prototype numbers with no comparison baselines.
Is the ATI code or dataset publicly available?
No. The paper does not mention any public code release, workload scripts, or dataset. The benchmark tool (wrk) is open source, but the ATI prototype itself, its resolver index implementation, and the tested metadata are not stated to be available, which limits independent reproduction.
About this review
This is an AI-generated preliminary review by Evolve.Pub, produced from the openly available text of the paper. It is not peer review and does not represent the authors' views. The canonical record of the paper is at the publisher link above.
- Figures 1-4 are absent from the extracted text; architecture and certificate-flow assessments rely on captions and body prose only.
- The extraction contains typographic corruption ('95th %lie', 'AgnentDNS'); all numbers are taken as reported.
- No appendix or supplementary material was visible; workload details may exist outside the provided text.
Go deeper on this paper
Wrote a paper? Stress-test it before reviewers do.
Upload your manuscript for the same structured scrutiny — reasoning, statistics, evidence, figures, and reproducibility.