English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

In targeted poisoning attacks, an attacker manipulates an agent-environmentinteraction to force the agent into adopting a policy of interest, calledtarget policy. Prior work has primarily focused on attacks that modify standardMDP primitives, such as rewards or transitions. In this paper, we studytargeted poisoning attacks in a two-agent setting where an attacker implicitlypoisons the effective environment of one of the agents by modifying the policyof its peer. We develop an optimization framework for designing optimalattacks, where the cost of the attack measures how much the solution deviatesfrom the assumed default policy of the peer agent. We further study thecomputational properties of this optimization framework. Focusing on a tabularsetting, we show that in contrast to poisoning attacks based on MDP primitives(transitions and (unbounded) rewards), which are always feasible, it is NP-hardto determine the feasibility of implicit poisoning attacks. We providecharacterization results that establish sufficient conditions for thefeasibility of the attack problem, as well as an upper and a lower bound on theoptimal cost of the attack. We propose two algorithmic approaches for findingan optimal adversarial policy: a model-based approach with tabular policies anda model-free approach with parametric/neural policies. We showcase the efficacyof the proposed algorithms through experiments.

Implicit Poisoning Attacks in Two-Agent Reinforcement Learning: Adversarial Policies for Training-Time Attacks