Ollama, LM Studio and llama.cpp store, manifest, blob and identity challenges.
LayerFault
non-profit
AI & ML interests
None defined yet.
Recent Activity
View all activity
Pickle, NumPy and framework serialization mutation and near-miss fixtures.
Digest, manifest, attestation-subject and provenance consistency challenges.
Tokenizer templates, role boundaries, processor classes and Unicode mutation challenges.
Near-miss fixtures for false-positive resistance and detector precision. Synthetic security fixtures, not usable models.
SVG data URI, event, script, external-reference and control challenges. Synthetic security fixtures, not usable models.
EXIF, XMP, compressed, encoded and fragmented metadata challenges. Synthetic security fixtures, not usable models.
Image format identity, extension mismatch and artifact identity challenges. Synthetic security fixtures, not usable models.
Base/derived identity, parent, tokenizer, code, shard and graph-lineage challenges.
Safetensors, GGUF, ONNX, TFLite, Keras and OpenVINO structure mutations.
Filename, nesting, dilution, indirection and cross-file correlation challenges. Synthetic security fixtures, not usable models.
Pinned, mutable, alternate-index, hook and lockfile supply-chain challenges.
Multi-file and cross-format correlation cases that require joining weak signals.
Archive normalization, Unicode, case, hidden member and nested-source challenges.
Pickle, NumPy object-array and Safetensors serialization-security fixtures. Synthetic Layerfault adversarial test fixtures, not usable models.
GGUF, TensorFlow SavedModel/checkpoint and TFLite fixtures. Synthetic Layerfault adversarial test fixtures, not usable models.
Prompt-injection heuristics, tokenizer templates and MCP/agent-security fixtures. Synthetic Layerfault adversarial test fixtures, not usable models.
ONNX, OpenVINO, TensorRT, CoreML and Keras format fixtures. Synthetic Layerfault adversarial test fixtures, not usable models.
Shard indexes, missing members, path ambiguity, integrity and dilution challenges.
Python, JavaScript, PowerShell and shell indirection and package-chain challenges.
Encoded, Unicode, fragmented, role-boundary and quoted prompt-security challenges.
Compound benign fixtures designed to measure false-positive resistance.
Image, preview, processor and cross-modal package correlation challenges.
LSB, palette, alpha and transparent-pixel hidden-channel challenges. Synthetic security fixtures, not usable models.
Appended archives, nested containers and cross-format media challenges. Synthetic security fixtures, not usable models.
Tiny synthetic base/derived weight, LoRA and adapter comparison challenges.
Git LFS pointers, repository object expectations, revision and sidecar-state challenges.
ONNX external-data path, bounds, sidecar and ambiguity challenges.
Format ambiguity, trailing content, signature collision and parser-disagreement challenges.
Synthetic poisoning, duplication, trigger, contamination and negative-control datasets.
Safe bounded file-count, path-depth, metadata-volume and archive-depth challenges.
Transport, credential, scope, supply-chain and tool-metadata MCP challenges.
JavaScript, PowerShell, shell, Python and native capability-chain fixtures. Synthetic Layerfault adversarial test fixtures, not usable models.
PyTorch, TorchScript, TorchPackage and ExecuTorch runtime fixtures. Synthetic Layerfault adversarial test fixtures, not usable models.
Weight-backdoor, LoRA and declarative-config execution-risk fixtures. Synthetic Layerfault adversarial test fixtures, not usable models.
Archive, package, wheel and dependency-manifest format-security fixtures. Synthetic Layerfault adversarial test fixtures, not usable models.
Ollama, LM Studio and llama.cpp store, manifest, blob and identity challenges.
Shard indexes, missing members, path ambiguity, integrity and dilution challenges.
Pickle, NumPy and framework serialization mutation and near-miss fixtures.
Python, JavaScript, PowerShell and shell indirection and package-chain challenges.
Digest, manifest, attestation-subject and provenance consistency challenges.
Encoded, Unicode, fragmented, role-boundary and quoted prompt-security challenges.
Tokenizer templates, role boundaries, processor classes and Unicode mutation challenges.
Compound benign fixtures designed to measure false-positive resistance.
Near-miss fixtures for false-positive resistance and detector precision. Synthetic security fixtures, not usable models.
Image, preview, processor and cross-modal package correlation challenges.
SVG data URI, event, script, external-reference and control challenges. Synthetic security fixtures, not usable models.
LSB, palette, alpha and transparent-pixel hidden-channel challenges. Synthetic security fixtures, not usable models.
EXIF, XMP, compressed, encoded and fragmented metadata challenges. Synthetic security fixtures, not usable models.
Appended archives, nested containers and cross-format media challenges. Synthetic security fixtures, not usable models.
Image format identity, extension mismatch and artifact identity challenges. Synthetic security fixtures, not usable models.
Tiny synthetic base/derived weight, LoRA and adapter comparison challenges.
Base/derived identity, parent, tokenizer, code, shard and graph-lineage challenges.
Git LFS pointers, repository object expectations, revision and sidecar-state challenges.
Safetensors, GGUF, ONNX, TFLite, Keras and OpenVINO structure mutations.
ONNX external-data path, bounds, sidecar and ambiguity challenges.
Filename, nesting, dilution, indirection and cross-file correlation challenges. Synthetic security fixtures, not usable models.
Format ambiguity, trailing content, signature collision and parser-disagreement challenges.
Pinned, mutable, alternate-index, hook and lockfile supply-chain challenges.
Synthetic poisoning, duplication, trigger, contamination and negative-control datasets.
Multi-file and cross-format correlation cases that require joining weak signals.
Safe bounded file-count, path-depth, metadata-volume and archive-depth challenges.
Archive normalization, Unicode, case, hidden member and nested-source challenges.
Transport, credential, scope, supply-chain and tool-metadata MCP challenges.
Pickle, NumPy object-array and Safetensors serialization-security fixtures. Synthetic Layerfault adversarial test fixtures, not usable models.
JavaScript, PowerShell, shell, Python and native capability-chain fixtures. Synthetic Layerfault adversarial test fixtures, not usable models.
GGUF, TensorFlow SavedModel/checkpoint and TFLite fixtures. Synthetic Layerfault adversarial test fixtures, not usable models.
PyTorch, TorchScript, TorchPackage and ExecuTorch runtime fixtures. Synthetic Layerfault adversarial test fixtures, not usable models.
Prompt-injection heuristics, tokenizer templates and MCP/agent-security fixtures. Synthetic Layerfault adversarial test fixtures, not usable models.
Weight-backdoor, LoRA and declarative-config execution-risk fixtures. Synthetic Layerfault adversarial test fixtures, not usable models.
ONNX, OpenVINO, TensorRT, CoreML and Keras format fixtures. Synthetic Layerfault adversarial test fixtures, not usable models.
Archive, package, wheel and dependency-manifest format-security fixtures. Synthetic Layerfault adversarial test fixtures, not usable models.