Abstract Recent advancements in deep neural networks (DNNs) have made them indispensable for numerous commercial applications. These include healthcare systems and self-driving cars. Training DNN models typically demands substantial time, vast datasets and high computational costs. However, these valuable models face significant risks. Attackers can steal and sell pre-trained DNN models for profit. Unauthorised sharing of these models poses a serious threat. Once sold, they can be easily copied and redistributed. Therefore, a well-built pre-trained DNN model is a valuable asset that requires protection. This paper introduces a robust hybrid two-level protection system for safeguarding the ownership of pre-trained DNN models. The first-level employs zero-bit watermarking. The second-level incorporates an adversarial attack as a watermark by using a perturbation technique to embed the watermark. The robustness of the proposed system is evaluated against seven types of attacks. These are Fast Gradient Method Attack, Auto Projected Gradient Descent Attack, Auto Conjugate Gradient Attack, Basic Iterative Method Attack, Momentum Iterative Method Attack, Square Attack and Auto Attack. The proposed two-level protection system withstands all seven attack types. It maintains accuracy and surpasses current state-of-the-art methods.

Nowadays, deep learning achieves higher levels of accuracy than ever before. This evolution makes deep learning crucial for applications that care for safety, like self-driving cars and helps consumers to meet most of their expectations. Further, Deep Neural Networks (DNNs) are powerful approaches that employed to solve several issues. These issues include healthcare, advertising, marketing, computer vision, speech processing, natural language processing. The DNNs have marvelous progress in these different fields, but training such DNN models requires a lot of time, a vast amount of data and in most cases a lot of computational steps. Selling such pre-trained models is a profitable business model. But, sharing them without the owner permission is a serious threat. Unfortunately, once the models are sold, they can be easily copied and redistributed. This paper first presents a review of how digital watermarking technologies are really very helpful in the copyright protection of the DNNs. Then, a comparative study between the latest techniques is presented. Also, several optimizers are proposed to improve the accuracy against the fine-tuning attack. Finally, several experiments are performed with black-box settings using several optimizers and the results are compared with the SGD optimizer.

A finite-state machine (FSM) is an abstract mathematical model of computation used to design both computer programs and sequential logic circuits. Considered as an abstract model of computation, FSM is weak; it has less computational power than some other models of computation such as the Turing machine. This paper discusses the finite-state automata based on Deoxyribonucleic Acid (DNA) and different implementations of DNA FSMs. Moreover, a comparison was made to clarify the advantages and disadvantages of each kind of presented DNA FSMS. Since it is a major goal for nanoscince, nanotechnology and super molecular chemistry is to design synthetic molecular devices that are programmable and run autonomously. Programmable means that the behavior of the device can be modified without redesigning the whole structure. Autonomous means that it runs without externally mediated change to the work cycle. In this paper we present an odd Parity Checker Prototype Using DNAzyme FSM. Our paper makes use of a known design for a DNA nanorobotic device due to Reif and Sahu for executing FSM computations using DNAzymes. The main contribution of our paper is a description of how to program that device to do a FSM computation known as odd parity checking. We describe in detail finite state automaton built on 10-23 DNAzyme, and give its procedure of design and computation. The design procedure has two major phases: designing the language potential alphabet DNA strands, and depending on the first phase to design the DNAzyme possible transitions.

We study the impact of cheating nodes in application-level multicast overlay trees. We focus on selfish nodes acting independently, cheating about their distance measurements during the control phase building or maintaining the tree. More precisely, we study, through simulations, the impact of simple cheating strategies in four protocols, representatives of different application-level multicast protocol "families": HBM (a protocol based on a centralized approach), TBCP (a distributed, tree first protocol), NICE (a distributed, tree first protocol based on clustering) and NARADA (a mesh first protocol). We evaluate the impact of cheats on the performance of the overlay trees as perceived by their nodes and the underlying network.

This paper considers an application-level multicast protocol, HBM, which can be used when native multicast routing is not available. Being purely end-to-end, application-level multicast proposals in general, and HBM in particular, are intrinsically more fragile than traditional routing solutions relying on well administered and dedicated routers. Improving their robustness is therefore of high practical importance and we believe it is a key aspect for the acceptance of the technology by end users who won't tolerate that a multi-participant video-conference session be subject to frequent cuts. In this work we identify two classes of problems that lead to packet losses, and for each class we introduce and compare several schemes. Experiments show that in both cases simple yet efficient solutions exist.

In this paper we show how to build a fully secure and efficient group communication service between several sites. This service is built on top of a VPN environment where IPSec tunnels are created, on-demand, between the various sites that need to communicate. This paper is a follow-up of previous work on group communications in a VPN environment and on application-level multicast. We show that these proposal naturally fit with one-another and lead to the concept of virtual private routed network or VPRN. This concept enables us to largely improve the data distribution efficiency, and in particularly reduces the physical link stress. We are convinced that security is critical in many situations and must be the primary concern of a group communication service.