LLMmap active probing remains useful diagnostic context, but it is not a scored dimension in the current model. The shipped implementation is a lexical / structural heuristic, not the trained contrastive classifier from the paper.
Algorithm
Send the enabled LLMmap-family probes, extract refusal-template, hedging, structure, and signature-token features, and compare them with coarse vendor templates. Return Unknown when coverage is too low or the top template does not beat the runner-up by enough margin.
Thresholds
Condition
Verdict contribution
Sufficient probe coverage and clear top template
Report a diagnostic vendor guess only
Low coverage or weak margin
Report Unknown
Any result
Score contribution remains 0
Limitations
System prompts and proxy rewriting can forge this signal. The implementation does not reproduce the paper's model, training data, or published accuracy.
References
Pasquini et al. LLMmap: Fingerprinting Large Language Models. USENIX Security 2025. arXiv:2407.15847
Anysingle signal cannot provemalicious behavior. Proxies may show anomalies for legitimate reasons (regional routing, A/B testing, degradation strategies, cache optimization).
Token ratio deviation may result from ChatML wrapping, system prompt injection, or tokenizer version differences — not necessarily intentional inflation.
Model identity judgment is based on statistical fingerprint matching, not cryptographic proof. Quantization, fine-tuning, and post-processing can all alter fingerprints.
MMD distribution tests are sensitive to temperature, sampling parameters, and system prompts. Significant p-values mean distributional difference, not proof of substitution.
Logprobs unavailability is increasingly common (many providers disable it by default in 2025-2026) and does not by itself indicate deception.
ITT rhythm fingerprinting is an early-stage technique. Network jitter, TCP coalescing, and gateway buffering can produce false signals.
This tool generates reference-grade evidence chains, not legal conclusions. Do not make definitive accusations based solely on this report.
The wording in the report refers to statistical "deviations" or "signal inconsistencies". Please do not use this to make fraud or deception claims against any service provider.