MétaCan
Menu
Back to cohort
Record W4400375395 · doi:10.48550/arxiv.2407.02551

Breach By A Thousand Leaks: Unsafe Information Leakage in `Safe' AI Responses

2024· preprint· en· W4400375395 on OpenAlexfundno aff
D. O. Glukhov, Ziwen Han, Ilia Shumailov, Vardan Papyan, Nicolas Papernot

Bibliographic record

VenuearXiv (Cornell University) · 2024
Typepreprint
Languageen
FieldComputer Science
TopicAdversarial Robustness in Machine Learning
Canadian institutionsnot available
FundersNatural Sciences and Engineering Research Council of CanadaGovernment of CanadaCanadian Institute for Advanced ResearchAlfred P. Sloan Foundation
KeywordsLeakage (economics)Computer securitySense (electronics)Computer scienceBusinessEngineeringElectrical engineeringEconomics

Abstract

fetched live from OpenAlex

Vulnerability of Frontier language models to misuse and jailbreaks has prompted the development of safety measures like filters and alignment training in an effort to ensure safety through robustness to adversarially crafted prompts. We assert that robustness is fundamentally insufficient for ensuring safety goals, and current defenses and evaluation methods fail to account for risks of dual-intent queries and their composition for malicious goals. To quantify these risks, we introduce a new safety evaluation framework based on impermissible information leakage of model outputs and demonstrate how our proposed question-decomposition attack can extract dangerous knowledge from a censored LLM more effectively than traditional jailbreaking. Underlying our proposed evaluation method is a novel information-theoretic threat model of inferential adversaries, distinguished from security adversaries, such as jailbreaks, in that success is measured by inferring impermissible knowledge from victim outputs as opposed to forcing explicitly impermissible outputs from the victim. Through our information-theoretic framework, we show that to ensure safety against inferential adversaries, defense mechanisms must ensure information censorship, bounding the leakage of impermissible information. However, we prove that such defenses inevitably incur a safety-utility trade-off.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.

How this classification was reachedexpand

Full frame machine prediction

Teacher imitation

Not calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.

metaresearch head score (Codex)0.022
metaresearch head score (Gemma)0.099
Version: metacan-v3-hybrid-931329e0061cValidation status: machine_predicted_unvalidated
Candidate categoriesnone
Consensus categoriesnone
DomainCandidate signal: none · Consensus signal: none
Study designCandidate signal: Bench or experimental · Consensus signal: none
GenreCandidate signal: Empirical · Consensus signal: none
Teacher disagreement score0.022
Threshold uncertainty score0.115

Distilled classifier scores by category (both heads)

CategoryCodexGemma
Metaresearch0.0220.099
Meta-epidemiology (narrow)0.0010.001
Meta-epidemiology (broad)0.0010.001
Bibliometrics0.0010.001
Science and technology studies0.0010.006
Scholarly communication0.0040.010
Open science0.0020.007
Research integrity0.0030.006
Insufficient payload (model declined to judge)0.0040.001

Machine scores (provisional)

The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.

Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.

Opus teacher head0.029
GPT teacher head0.206
Teacher spread0.177 · how far apart the two teachers sit on this one work
Validation statusscore_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it

Classification

machine, unvalidated

Machine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.

The models applied no category: nothing in the taxonomy fit this work.
Study designBench or experimental
Domainnot available
GenreEmpirical

How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".

Quick stats

Citations0
Published2024
Admission routes1
Has abstractyes

Explore more

Same venuearXiv (Cornell University)Same topicAdversarial Robustness in Machine LearningFrench-language works237,207