Paper

SoK: Colluding Adversaries in Machine Learning Pipelines

Unreadunread

AI summary

arXiv:2606. 10091v1 Announce Type: new Abstract: Machine learning (ML) models are susceptible to various security, privacy, and fairness risks.

Reading status

Priority

About this paper

Authors: Vasisht Duddu, Lipeng He, Asim Waheed, N. Asokan
Released: 2026-06-10 (29d ago)
Topic: Related to current project
Relation to my work: Read next because SoK: Colluding Adversaries in Machine Learning Pipelines overlaps with clean result "The marker is a representational handle, not a behavioural one — sharing it between a villain persona and the assistant transfers no misalignment (HIGH confidence)", experiment "Factor screen for marker implantation + leakage (2^5: system-prompt length, answer-format length, persona-presence, on-policy, marker-only-loss)", experiment "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: line, factor, model. Source: arxiv cs.CR (Cryptography and Security).
arXiv: 2606.10091

Ask Claude about this paper

Posts a comment with @claude so the runner spins up a Claude Code session scoped to SoK: Colluding Adversaries in Machine Learning Pipelines. Replies appear in the conversation below.

Comments

No comments yet.

Related records

No links yet.