Published on Sun Jan 17 2021

Adversarial Attacks On Multi-Agent Communication

James Tu, Tsunhsuan Wang, Jingkang Wang, Sivabalan Manivasagam, Mengye Ren, Raquel Urtasun

Modern autonomous systems will soon be deployed at scale, opening up the possibility for cooperative multi-agent systems. By sharing information and distributing workloads, autonomous agents can better perform their tasks. Suchadvantages rely heavily on communication channels which have been shown to be vulnerable to security breaches.

0
0
0
Abstract

Growing at a very fast pace, modern autonomous systems will soon be deployed at scale, opening up the possibility for cooperative multi-agent systems. By sharing information and distributing workloads, autonomous agents can better perform their tasks and enjoy improved computation efficiency. However, such advantages rely heavily on communication channels which have been shown to be vulnerable to security breaches. Thus, communication can be compromised to execute adversarial attacks on deep learning models which are widely employed in modern systems. In this paper, we explore such adversarial attacks in a novel multi-agent setting where agents communicate by sharing learned intermediate representations. We observe that an indistinguishable adversarial message can severely degrade performance, but becomes weaker as the number of benign agents increase. Furthermore, we show that transfer attacks are more difficult in this setting when compared to directly perturbing the inputs, as it is necessary to align the distribution of communication messages with domain adaptation. Finally, we show that low-budget online attacks can be achieved by exploiting the temporal consistency of streaming sensory inputs.

Sat May 21 2016
Artificial Intelligence
Learning to Communicate with Deep Multi-Agent Reinforcement Learning
We consider the problem of multiple agents sensing and acting in environments. We propose two approaches for learning in these domains. The former uses deep Q-learning, while the latter exploits the fact that agents can backpropagate error derivatives through noisy communication channels.
0
0
0
Tue Mar 12 2019
Artificial Intelligence
On the Pitfalls of Measuring Emergent Communication
The majority of recent papers on emergent communication show that adding acommunication channel leads to an increase in reward or task success. This is a useful indicator, but provides only a coarse measure of the agent's learned communication abilities. In this paper, we examine a few intuitive existing metrics for measuring
0
0
0
Sat Jan 12 2019
Artificial Intelligence
Improving Coordination in Small-Scale Multi-Agent Deep Reinforcement Learning through Memory-driven Communication
Deep reinforcement learning algorithms have recently been used to train multiple interacting agents in a centralised manner. When the agents can only acquire partial observations and are faced with tasks requiring coordination and synchronisation skills,inter-agent communication plays an essential role.
0
0
0
Mon Dec 03 2018
Machine Learning
Multi-agent Deep Reinforcement Learning with Extremely Noisy Observations
Multi-agent reinforcement learning systems aim to provide interacting agents with the ability to collaboratively learn and adapt to the behaviour of other agents. In many real-world applications, the agents can only acquire a partial view of the world. We propose a multi-agent deep deterministic policy gradient algorithm
0
0
0
Tue Dec 03 2019
Artificial Intelligence
Learning Agent Communication under Limited Bandwidth by Message Pruning
Communication is a crucial factor for the big multi-agent world to stay organized and productive. Many methods keep sending messages incessantly, which consumes too much bandwidth. To handle this problem, we propose a gating mechanism to adaptively prune less beneficial messages.
0
0
0
Fri Oct 26 2018
Artificial Intelligence
TarMAC: Targeted Multi-Agent Communication
We propose a targeted communication architecture for multi-agent learning. Agents learn both what messages to send and whom to address them to. This targeting behavior is learnt solely from downstream task-specific reward. We additionally augment this with a multi-round communication approach.
0
0
0