What's Pulling the Strings? Evaluating Integrity and Attribution in AI Training and Inference through Concept Shift
The growing adoption of artificial intelligence (AI) has amplified concerns about trustworthiness, including integrity, privacy, robustness, and bias. To assess and attribute these threats, we propose ConceptLens, a generic framework that leverages pre-trained multimodal models to identify the root causes of integrity threats by analyzing Concept Shift in probing samples. ConceptLens demonstrates strong detection performance for vanilla data poisoning attacks and uncovers vulnerabilities to bias injection, such as the generation of covert advertisements through malicious concept shifts. It identifies privacy risks in unaltered but high-risk samples, filters them before training, and provides insights into model weaknesses arising from incomplete or imbalanced training data. Additionally, at the model level, it attributes concepts that the target model is overly dependent on, identifies misleading concepts, and explains how disrupting key concepts negatively impacts the model. Furthermore, it uncovers sociological biases in generative content, revealing disparities across sociological contexts. Strikingly, ConceptLens reveals how safe training and inference data can be unintentionally and easily exploited, potentially undermining safety alignment. Our study informs actionable insights to breed trust in AI systems, thereby speeding adoption and driving greater innovation.
Code (0)
등록된 구현이 없습니다.
Tasks
AttributeData PoisoningSafety AlignmentSimilar Papers 제목 키워드 기반
Impact of interpersonal influences on Employee engagement and Psychological contract: Effects of guanxi, wasta, jeitinho, blat and pulling strings
This study puts forward a conceptual model linking interpersonal influences' impact on Employee Engagement, Psychological contracts, and Human Resource Practices. It builds on human and social capital, as well as the soc…
I Can Find You in Seconds! Leveraging Large Language Models for Code Authorship Attribution
Source code authorship attribution is important in software forensics, plagiarism detection, and protecting software patch integrity. Existing techniques often rely on supervised machine learning, which struggles with ge…
Adversarial RobustnessAttributeAuthorship AttributionFew-Shot Learning+1What Do Deepfake Speech Detectors Actually Hear?
Deepfake speech detectors often output a single score without explaining why an audio sample is flagged, where in the signal the evidence lies, or what cues drive the decision. We propose an audio-native explainability p…
Ideal Attribution and Faithful Watermarks for Language Models
We introduce ideal attribution mechanisms, a formal abstraction for reasoning about attribution decisions over strings. At the core of this abstraction lies the ledger, an append-only log of the prompt-response interacti…
The Attribution Contract: Feature Attribution for Generative Language Models
Feature attribution methods promise to identify which input features matter for a model output. In generative language models, however, it is often unclear what should count as a feature in the first place. In autoregres…