REVIEW 11 cited by
Universal Zero-shot Embedding Inversion
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
Universal Zero-shot Embedding Inversion
read the original abstract
Embedding inversion, i.e., reconstructing text given its embedding and black-box access to the embedding encoder, is a fundamental problem in both NLP and security. From the NLP perspective, it helps determine how much semantic information about the input is retained in the embedding. From the security perspective, it measures how much information is leaked by vector databases and embedding-based retrieval systems. State-of-the-art methods for embedding inversion, such as vec2text, have high accuracy but require (a) training a separate model for each embedding, and (b) a large number of queries to the corresponding encoder. We design, implement, and evaluate ZSInvert, a zero-shot inversion method based on the recently proposed adversarial decoding technique. ZSInvert is fast, query-efficient, and can be used for any text embedding without training an embedding-specific inversion model. We measure the effectiveness of ZSInvert on several embeddings and demonstrate that it recovers key semantic information about the corresponding texts.
Forward citations
Cited by 11 Pith papers
-
Embedding Inference Attack
Tailored queries enable identification of the embedding model used by a black-box IR system from the unordered set of retrieved documents, even when a reranker is present.
-
SHARD: cell-keyed residual splitting for alignment-resistant private dense retrieval
SHARD shards private embedding residuals into cell-local keyed groups to raise the anchor requirement for alignment attacks by a factor of C while preserving full-dimensional nDCG@10 via encrypted reranking.
-
Can You Trust the Vectors in Your Vector Database? Black-Hole Attack from Embedding Space Defects
Injecting a few malicious vectors near the centroid exploits centrality-driven hubness in high-dimensional embeddings, causing them to dominate top-k retrievals in up to 99.85% of cases.
-
SHARD: cell-keyed residual splitting for alignment-resistant private dense retrieval
Cell-keyed residual splitting recovers raw dense-retrieval nDCG while raising diffuse known-plaintext alignment cost roughly C-fold, without cryptographic document privacy or full unlinkability.
-
SHARD: cell-keyed residual splitting for alignment-resistant private dense retrieval
SHARD introduces cell-keyed residual splitting that turns dense retrieval embeddings into revocable, renewable, unlinkable templates resistant to alignment attacks while preserving exact utility under CKKS reranking.
-
Hybrid privacy-aware semantic search: SVD-truncated document geometry and CKKS-encrypted query reranking under a restricted threat model
A hybrid scheme combines SVD-truncated and rotated document vectors with CKKS-encrypted query reranking to preserve semantic search ranking quality at million-document scale under a restricted threat model while resis...
-
Towards Privacy-Preserving Large Language Model: Text-free Inference Through Alignment and Adaptation
PPFT enables text-free LLM inference by aligning client embeddings with server projection and LLM, then fine-tuning on noise-injected private embeddings to maintain near-baseline performance without exposing raw prompts.
-
Can You Trust the Vectors in Your Vector Database? Black-Hole Attack from Embedding Space Defects
Injecting a few vectors near the embedding-space centroid can make them appear in top-k results for up to 94.4% of queries via centrality-driven hubness.
-
Hybrid privacy-aware semantic search: SVD-truncated document geometry and CKKS-encrypted query reranking under a restricted threat model
SVD truncation plus a secret rotation and CKKS query reranking can hide numerical query slots at sub-second latency on 1M docs, but document privacy is only empirical obfuscation that fails with about k known pairs.
-
Hybrid privacy-aware semantic search: SVD-truncated document geometry and CKKS-encrypted query reranking under a restricted threat model
SVD truncation plus a secret rotation on documents and CKKS on queries preserves sub-second retrieval quality while collapsing off-the-shelf inversion, with document protection failing under known-plaintext Procrustes...
-
Hybrid privacy-aware semantic search: SVD-truncated document geometry and CKKS-encrypted query reranking under a restricted threat model
Hybrid privacy method for semantic search truncates and rotates document vectors geometrically while encrypting queries with CKKS, preserving retrieval quality on 1M-document corpora under a restricted threat model.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.