all InfoSec news
Spinning Language Models: Risks of Propaganda-As-A-Service and Countermeasures. (arXiv:2112.05224v2 [cs.CR] UPDATED)
April 11, 2022, 1:20 a.m. | Eugene Bagdasaryan, Vitaly Shmatikov
cs.CR updates on arXiv.org arxiv.org
We investigate a new threat to neural sequence-to-sequence (seq2seq) models:
training-time attacks that cause models to "spin" their outputs so as to
support an adversary-chosen sentiment or point of view -- but only when the
input contains adversary-chosen trigger words. For example, a spinned
summarization model outputs positive summaries of any text that mentions the
name of some individual or organization.
Model spinning introduces a "meta-backdoor" into a model. Whereas
conventional backdoors cause models to produce incorrect outputs on inputs …
More from arxiv.org / cs.CR updates on arXiv.org
Jobs in InfoSec / Cybersecurity
Information Technology Specialist II: Network Architect
@ Los Angeles County Employees Retirement Association (LACERA) | Pasadena, CA
Cybersecurity Skills Challenge -- Sponsored by DoD
@ Correlation One | United States
Security Operations Center (SOC) Analyst
@ GK Cybersecurity Group | Remote
Lead Product Security Engineer
@ Baker Hughes | IN-KA-BANGALORE-NEON BUILDING WEST TOWER
Penetration Tester
@ BT Group | Hemel Hempstead: Riverside (R6, Hemel Hempstead, United Kingdom
Cloud and Infrastructure Security Engineer II
@ StubHub | Los Angeles, CA