Thomas Perry Jr. · SignaBuilder · ORCID 0009-0007-1476-1213
Security research on synthetic identity, verification infrastructure, and autonomous detection systems.
Presents the Agent Delegated Purchasing Protocol (ADPP), a framework for trust-scored autonomous commerce where AI agents execute purchasing decisions on behalf of principals with cryptographic accountability and real-time trust verification.
Documents a real-world firmware-persistent macOS compromise involving steganographic command channels, firmware-level persistence mechanisms, and active defense countermeasures. Provides forensic methodology and defensive steganography techniques for incident responders.
The fabrication of credible identity is now an industrial process. Generative AI has compressed the cost of producing synthetic persons, institutions, and authority signals to near zero, while verifying authenticity remains structurally expensive. This paper presents a verification infrastructure framework for the synthetic era.
Controlled self-referential experiment in synthetic authority construction. Maps escalation from media deepfakes to full synthetic institutions. Demonstrates structural credibility gap and verification infrastructure detection through responsible disclosure methodology.
This paper formalizes the concept of entity-level deepfakes—synthetic constructs that maintain persistent digital identities across platforms, participate in fabricated reference networks, and are designed to be indistinguishable from authentic entities. Introduces the Synthetic Density Index and proposes verified training corpus frameworks.
Demonstrates that a complete organizational impersonation infrastructure—fabricated company, deepfake-capable executives, compromised AI model contamination, and customer-facing operations—can be established for under $500 initial cost. The attack exploits three convergent vectors: real-time deepfake generation, recursive AI training data contamination, and the governance gap affecting 73% of organizations. Presents Helix Fabric v2 with eight independent detection vectors achieving greater than 0.85 detection confidence.
Demonstrates that a fully believable synthetic organization with executive personas and publication histories can be fabricated for approximately $50 initial costs. Presents Helix Fabric v2, a distributed detection framework with 15 signal types across multiple worker categories.
A serverless meta-orchestration system built on Cloudflare Workers that autonomously detects synthetic organizations, AI-generated personas, and intellectual property clones in the faith-technology ecosystem. Multi-signal detection across fleet-spawned workers.
A unified framework for anticipatory cyber defense integrating eight convergent dimensions: adversarial machine learning countermeasures, supply chain and hardware implant analysis, quantum threat transition analysis, attribution resistance with deepfake forensics, autonomous defense game theory, zero-knowledge proof systems for operational security, temporal correlation at scale, and biological-physical security integration.
Extended version (28 pages) with formal model (9 definitions, 3 theorems). Presents a pipeline parallelism model for multi-agent SaaS development using helical architecture, constitutional AI governance, and gradient auto-compaction for LLM agent orchestration.
Analyzes the 5Ghoul vulnerability family affecting 5G NR implementations and develops practical defensive countermeasures including firmware-level mitigations, network-side detection heuristics, and protocol-aware monitoring strategies for mobile infrastructure operators.
Presents the Transaction Trust Score (TTS), a real-time scoring system that aggregates four independent signals—entity verification, constraint compliance history, behavioral fingerprinting, and counterparty risk assessment—into a single trust verdict delivered in under 200ms. Formalized as the Delegated Agent Commerce Protocol (DACP), published on the Buf Schema Registry.
Argues that hardware-based trust (Trusted Execution Environments) and entity-based trust (behavioral verification scoring) are complementary attestation models. Formalizes a dual-attestation architecture integrating TEE hardware proofs with Transaction Trust Scoring (TTS). Valid TEE attestation improves trust scores by 20%, while invalid TEE collateral decreases confidence by 62%.
Presents a multi-LLM peer orchestration framework where heterogeneous language models collaborate through structured debate, constitutional governance, and consensus protocols to produce higher-quality outputs than any single model.
Formalizes constitutional governance constraints for autonomous AI systems, defining hard boundaries, soft preferences, and escalation protocols that ensure alignment without sacrificing capability.
Applies formal verification techniques using Coq proof assistant and Z3 SMT solver to prove safety properties of AI governance constraint systems, establishing machine-checkable guarantees for constitutional AI frameworks.
Establishes a methodology for constructing cryptographically verifiable prior art evidence chains using persistent identifiers, timestamped deposits, and hash-linked provenance records.
Presents an intellectual property evidence chain framework that links DOI-anchored publications, code repositories, and deployment artifacts into a tamper-evident provenance graph for defending IP claims.