Deep Learning (DL) algorithms based on artificial neural networks have achieved remarkable success and are being extensively applied in a variety of application domains, ranging from image classification, automatic driving, natural language processing to medical diagnosis, credit risk assessment, intrusion detection. However, the privacy and security issues of DL have been revealed that the DL model can be stolen or reverse engineered, sensitive training data can be inferred, even a recognizable face image of the victim can be recovered. Besides, the recent works have found that the DL model is vulnerable to adversarial examples perturbed by imperceptible noised, which can lead the DL model to predict wrongly with high confidence. In this paper, we first briefly introduces the four types of attacks and privacy-preserving techniques in DL. We then review and summarize the attack and defense methods associated with DL privacy and security in recent years. To demonstrate that security threats really exist in the real world, we also reviewed the adversarial attacks under the physical condition. Finally, we discuss current challenges and open problems regarding privacy and security issues in DL.

Big data, because it can mine new knowledge for economic growth and technical innovation, has recently received considerable attention, and many research efforts have been directed to big data processing due to its high volume, velocity, and variety (referred to as "3V") challenges. However, in addition to the 3V challenges, the flourishing of big data also hinges on fully understanding and managing newly arising security and privacy challenges. If data are not authentic, new mined knowledge will be unconvincing; while if privacy is not well addressed, people may be reluctant to share their data. Because security has been investigated as a new dimension, "veracity," in big data, in this article, we aim to exploit new challenges of big data in terms of privacy, and devote our attention toward efficient and privacy-preserving computing in the big data era. Specifically, we first formalize the general architecture of big data analytics, identify the corresponding privacy requirements, and introduce an efficient and privacy-preserving cosine similarity computing protocol as an example in response to data mining's efficiency and privacy requirements in the big data era.

In this paper, we propose a toolkit for efficient and privacy-preserving outsourced calculation under multiple encrypted keys (EPOM). Using EPOM, a large scale of users can securely outsource their data to a cloud server for storage. Moreover, encrypted data belonging to multiple users can be processed without compromising on the security of the individual user's (original) data and the final computed results. To reduce the associated key management cost and private key exposure risk in EPOM, we present a distributed two-trapdoor public-key cryptosystem, the core cryptographic primitive. We also present the toolkit to ensure that the commonly used integer operations can be securely handled across different encrypted domains. We then prove that the proposed EPOM achieves the goal of secure integer number processing without resulting in privacy leakage of data to unauthorized parties. Last, we demonstrate the utility and the efficiency of EPOM using simulations.

Due to resource constraints and working surroundings, many IIoT nodes are easily hacked and turn into zombies from which to launch attacks. It is challenging to detect such networked zombies rooted behind the Internet for any individual defender. In this article, we combine federated learning (FL) and fog/edge computing to combat malicious codes. Our protocol trains a global optimized model based on distributed datasets of collaborators while removing the data and communication constraints. The FL-based detection protocol maximizes the values of distributed data samples, resulting in an accurate model timely. On top of the protocol, we place mitigation intelligence in a distributed and collaborative manner. Our approach improves accuracy, eliminates mitigation time, and enlarges attackers’ expense within a defense alliance. Comprehensive evaluations confirm that the cost incurred is 2.7 times larger, the mitigation response time is 72% lower, and the accuracy is 47% higher on average. Besides, the protocol evaluation shows the detection accuracy is approximately 98% in the FL, which is almost the same as centralized training.