To address the costs of reverse engineering and signature extraction, advanced research on malware detection focuses on using neural networks to learn malicious behaviors with static and dynamic features. However, deep learning-based malware detection models are vulnerable to a hack from adversarial samples. The attackers' goal is to generate imperceptible perturbations to the original samples and evade detection. In the context of malware, the generated samples should have one more important character: it should not change the malicious behaviors of the original code. So the original features can not be removed and changed. In this paper, we proposed a reinforcement learning based attack to deceive graph based malware detection models. Inspired by obfuscation techniques, the central idea of the proposed attack is to sequentially inject semantic Nops, which will not change the program's functionality, into CFGs(Control Flow Graph). Specifically, the Semantics-preserving Reinforcement Learning(SRL) Attack is to learn an RL agent to iteratively select the semantic Nops and insert them into basic blocks of the CFGs. Variants of obfuscation methods, hill-climbing methods, and gradient based algorithms are proposed: 1) Semantics-preserving Random Insertion(SRI) Attack: randomly inserting semantic Nops into basic blocks.; 2) Semantics-preserving Accumulated Insertion(SAI) Attack: declining certain random transformation according to the probability of the target class; 3) Semantics-preserving Gradient based Insertion(SGI) Attack: applying transformation on the original CFG in the direction of the gradient. We use real-world Windows programs to show that a family of Graph Neural Network models are vulnerable to these attacks. The best evasion rate of the benchmark attacks are 97% on the basic GCN model and 96% on DGCNN model. The SRL attack can achieve 100% on both models.

We present NavACL, a method of automatic curriculum learning tailored to the navigation task. NavACL is simple to train and efficiently selects relevant tasks using geometric features. In our experiments, deep reinforcement learning agents trained using NavACL in collision-free environments significantly outperform state-of-the-art agents trained with uniform sampling -- the current standard. Furthermore, our agents are able to navigate through unknown cluttered indoor environments to semantically-specified targets using only RGB images. Collision avoidance policies and frozen feature networks support transfer to unseen real-world environments, without any modification or retraining requirements. We evaluate our policies in simulation, and in the real world on a ground robot and a quadrotor drone. Videos of real-world results are available in the supplementary material

Path planning for mobile robots in large dynamic environments is a challenging problem, as the robots are required to efficiently reach their given goals while simultaneously avoiding potential conflicts with other robots or dynamic objects. In the presence of dynamic obstacles, traditional solutions usually employ re-planning strategies, which re-call a planning algorithm to search for an alternative path whenever the robot encounters a conflict. However, such re-planning strategies often cause unnecessary detours. To address this issue, we propose a learning-based technique that exploits environmental spatio-temporal information. Different from existing learning-based methods, we introduce a globally guided reinforcement learning approach (G2RL), which incorporates a novel reward structure that generalizes to arbitrary environments. We apply G2RL to solve the multi-robot path planning problem in a fully distributed reactive manner. We evaluate our method across different map types, obstacle densities, and the number of robots. Experimental results show that G2RL generalizes well, outperforming existing distributed methods, and performing very similarly to fully centralized state-of-the-art benchmarks.

Imitation learning in a high-dimensional environment is challenging. Most inverse reinforcement learning (IRL) methods fail to outperform the demonstrator in such a high-dimensional environment, e.g., Atari domain. To address this challenge, we propose a novel reward learning module to generate intrinsic reward signals via a generative model. Our generative method can perform better forward state transition and backward action encoding, which improves the module's dynamics modeling ability in the environment. Thus, our module provides the imitation agent both the intrinsic intention of the demonstrator and a better exploration ability, which is critical for the agent to outperform the demonstrator. Empirical results show that our method outperforms state-of-the-art IRL methods on multiple Atari games, even with one-life demonstration. Remarkably, our method achieves performance that is up to 5 times the performance of the demonstration.

Recent work in deep reinforcement learning (RL) has produced algorithms capable of mastering challenging games such as Go, chess, or shogi. In these works the RL agent directly observes the natural state of the game and controls that state directly with its actions. However, when humans play such games, they do not just reason about the moves but also interact with their physical environment. They understand the state of the game by looking at the physical board in front of them and modify it by manipulating pieces using touch and fine-grained motor control. Mastering complicated physical systems with abstract goals is a central challenge for artificial intelligence, but it remains out of reach for existing RL algorithms. To encourage progress towards this goal we introduce a set of physically embedded planning problems and make them publicly available. We embed challenging symbolic tasks (Sokoban, tic-tac-toe, and Go) in a physics engine to produce a set of tasks that require perception, reasoning, and motor control over long time horizons. Although existing RL algorithms can tackle the symbolic versions of these tasks, we find that they struggle to master even the simplest of their physically embedded counterparts. As a first step towards characterizing the space of solution to these tasks, we introduce a strong baseline that uses a pre-trained expert game player to provide hints in the abstract space to an RL agent's policy while training it on the full sensorimotor control task. The resulting agent solves many of the tasks, underlining the need for methods that bridge the gap between abstract planning and embodied control.

Our method learns complex behaviors by training offline from prior datasets (expert demonstrations, data from previous experiments, or random exploration data) and then fine-tuning quickly with online interaction. Robots trained with reinforcement learning (RL) have the potential to be used across a huge variety of challenging real world problems. To apply RL to a new problem, you typically set up the environment, define a reward function, and train the robot to solve the task by allowing it to explore the new environment from scratch. While this may eventually work, these "online" RL methods are data hungry and repeating this data inefficient process for every new problem makes it difficult to apply online RL to real world robotics problems. What if instead of repeating the data collection and learning process from scratch every time, we were able to reuse data across multiple problems or experiments?

Risk is 6 player game with significant randomness and a large game-tree complexity which poses a challenge to creating an agent to play the game effectively. Previous AIs focus on creating high-level handcrafted features determine agent decision making. In this project, I create D.A.D, A Risk agent using temporal difference reinforcement learning to train a Deep Neural Network including a Graph Convolutional Network to evaluate player positions. This is used in a game-tree to select optimal moves. This allows minimal handcrafting of knowledge into the AI, assuring input features are as low-level as possible to allow the network to extract useful and sophisticated features itself, even with the network starting from a random initialisation. I also tackle the issue of non-determinism in Risk by introducing a new method of interpreting attack moves necessary for the search. The result is an AI which wins 35% of the time versus 5 of best inbuilt AIs in Lux Delux, a Risk variant.

Prisoner's Dilemma mainly treat the choice to cooperate or defect as an atomic action. We propose to study online learning algorithm behavior in the Iterated Prisoner's Dilemma (IPD) game, where we explored the full spectrum of reinforcement learning agents: multi-armed bandits, contextual bandits and reinforcement learning. We have evaluate them based on a tournament of iterated prisoner's dilemma where multiple agents can compete in a sequential fashion. This allows us to analyze the dynamics of policies learned by multiple self-interested independent reward-driven agents, and also allows us study the capacity of these algorithms to fit the human behaviors. Results suggest that considering the current situation to make decision is the worst in this kind of social dilemma game. Multiples discoveries on online learning behaviors and clinical validations are stated.

The ability to exploit prior experience to solve novel problems rapidly is a hallmark of biological learning systems and of great practical importance for artificial ones. In the meta reinforcement learning literature much recent work has focused on the problem of optimizing the learning process itself. In this paper we study a complementary approach which is conceptually simple, general, modular and built on top of recent improvements in off-policy learning. The framework is inspired by ideas from the probabilistic inference literature and combines robust off-policy learning with a behavior prior, or default behavior that constrains the space of solutions and serves as a bias for exploration; as well as a representation for the value function, both of which are easily learned from a number of training tasks in a multi-task scenario. Our approach achieves competitive adaptation performance on hold-out tasks compared to meta reinforcement learning baselines and can scale to complex sparse-reward scenarios.

Counterfactual regret minimization (CFR) is a popular method to deal with decision-making problems of two-player zero-sum games with imperfect information. Unlike existing studies that mostly explore for solving larger scale problems or accelerating solution efficiency, we propose a framework, RLCFR, which aims at improving the generalization ability of the CFR method. In the RLCFR, the game strategy is solved by the CFR in a reinforcement learning framework. And the dynamic procedure of iterative interactive strategy updating is modeled as a Markov decision process (MDP). Our method, RLCFR, then learns a policy to select the appropriate way of regret updating in the process of iteration. In addition, a stepwise reward function is formulated to learn the action policy, which is proportional to how well the iteration strategy is at each step. Extensive experimental results on various games have shown that the generalization ability of our method is significantly improved compared with existing state-of-the-art methods.