Biblio

With the development of artificial intelligence, the need for data sharing is becoming more and more urgent. However, the existing data sharing methods can no longer fully meet the data sharing needs. Privacy breaches, lack of motivation and mutual distrust have become obstacles to data sharing. We design a privacy-preserving, decentralized data sharing method based on blockchain smart contracts, named PPDS. To protect data privacy, we transform the data sharing problem into a model sharing problem. This means that the data owner does not need to directly share the raw data, but the AI model trained with such data. The data requester and the data owner interact on the blockchain through a smart contract. The data owner trains the model with local data according to the requester's requirements. To fairly assess model quality, we set up several model evaluators to assess the validity of the model through voting. After the model is verified, the data owner who trained the model will receive reward in return through a smart contract. The sharing of the model avoids direct exposure of the raw data, and the reasonable incentive provides a motivation for the data owner to share the data. We describe the design and workflow of our PPDS, and analyze the security using formal verification technology, that is, we use Coloured Petri Nets (CPN) to build a formal model for our approach, proving its security through simulation execution and model checking. Finally, we demonstrate effectiveness of PPDS by developing a prototype with its corresponding case application.

The metal oxide arrester (MOA, shortly) is installed on the line side of the substation, which is the first line of defense for the overvoltage limitation of lightning intrusion wave. In order to deeply limit the switching overvoltage and cancel the closing resistance of the circuit breaker, the arrester is replaced by the controllable metal oxide arrester (CMOA, shortly) in the new technology. The controllable switch of CMOA can be mechanical switch or thyristor switch. Thyristor switches are sensitive to the current and current change rate (di/dt) under lightning intrusion wave. If the switch cannot withstand, appropriate protective measures must be taken to ensure the safe operation of the controllable switch under this working condition. The 1000kV West Beijing to Shijiazhuang UHV AC transmission and transformation expansion project is the first project of pilot application of CMOA. CMOA were installed at both ends of the outgoing branch of Dingtai line I. In order to study the influence of lightning intrusion wave on the controllable switch of CMOA, this paper selected this project to simulate the lightning stroke on the incoming section of Dingtai line I in Beijing West substation in the process of system air closing or single-phase reclosing, and obtained the current and di/dt of the controllable switch through CMOA under this working condition. Then the performances of mechanical and thyristor control switches were checked respectively. The results showed that the mechanical switch could withstand without protective measures. The tolerance of thyristor switch to i and di/dt exceeded the limit value, and measures should be taken to protect and limit it. In this paper, the protection measures of current limiting reactor were given, and the limiting effect of the protection measures was verified by simulation and test. It could fully meet the requirements and ensure the safe operation of thyristor controllable switch.

Deep Neural Networks (DNN) has gained great success in solving several challenging problems in recent years. It is well known that training a DNN model from scratch requires a lot of data and computational resources. However, using a pre-trained model directly or using it to initialize weights cost less time and often gets better results. Therefore, well pre-trained DNN models are valuable intellectual property that we should protect. In this work, we propose DeepTrace, a framework for model owners to secretly fingerprinting the target DNN model using a special trigger set and verifying from outputs. An embedded fingerprint can be extracted to uniquely identify the information of model owner and authorized users. Our framework benefits from both white-box and black-box verification, which makes it useful whether we know the model details or not. We evaluate the performance of DeepTrace on two different datasets, with different DNN architectures. Our experiment shows that, with the advantages of combining white-box and black-box verification, our framework has very little effect on model accuracy, and is robust against different model modifications. It also consumes very little computing resources when extracting fingerprint.

We study the use of Multiform Logical Time, as embodied in Esterel/SyncCharts and Clock Constraint Specification Language (CCSL), for the specification of assume-guarantee constraints providing safe driving rules related to time and space, in the context of Automated Driving Assistance Systems (ADAS). The main novelty lies in the use of logical clocks to represent the epochs of specific area encounters (when particular area trajectories just start overlapping for instance), thereby combining time and space constraints by CCSL to build safe driving rules specification. We propose the safe specification pattern at high-level that provide the required expressiveness for safe driving rules specification. In the pattern, multiform logical time provides the power of parameterization to express safe driving rules, before instantiation in further simulation contexts. We present an efficient way to irregularly update the constraints in the specification due to the context changes, where elements (other cars, road sections, traffic signs) may dynamically enter and exit the scene. In this way, we add constraints for the new elements and remove the constraints related to the disappearing elements rather than rebuild everything. The multi-lane highway scenario is used to illustrate how to irregularly and efficiently update the constraints in the specification while receiving a fresh scene.

A number of solutions have been proposed to tackle the user privacy-preserving issue. Most of existing schemes, however, focus on methodology and techniques from the perspective of data processing. In this paper, we propose a lightweight privacy-preserving scheme for user identity from the perspective of data user and applied cryptography. The basic idea is to break the association relationships between User identity and his behaviors and ensure that User can access data or services as usual while the real identity will not be revealed. To this end, an interactive zero-knowledge proof protocol of identity is executed between CSP and User. Besides, a trusted third-party is introduced to manage user information, help CSP to validate User identity and establish secure channel between CSP and User via random shared key. After passing identity validation, User can log into cloud platform as usual without changing existing business process using random temporary account and password generated by CSP and sent to User by the secure channel which can further obscure the association relationships between identity and behaviors. Formal security analysis and theoretic and experimental evaluations are conducted, showing that the proposal is efficient and practical.

We present a formal OpenFlow-based network programming language (OF) including various flow rules, which can not only describe the behaviors of an individual switch, but also support to model a network of switches connected in the point-to-point topology. Besides, a topology-oriented operational semantics of the proposed language is explored to specify how the packet is processed and delivered in the OpenFlow-based networks. Based on the formal framework, we also propose an approach to detect potential security threats caused by the conflict of dynamic flow rules imposed by dynamic OpenFlow applications.

This paper proposed a new detection and prevention system against DDoS (Distributed Denial of Service) attack in SDN (software defined network) architecture, FL-GUARD (Floodlight-based guard system). Based on characteristics of SDN and centralized control, etc., FL-GUARD applies dynamic IP address binding to solve the problem of IP spoofing, and uses 3.3.2 C-SVM algorithm to detect attacks, and finally take advantage of the centralized control of software-defined network to issue flow tables to block attacks at the source port. The experiment results show the effectiveness of our system. The modular design of FL-GUARD lays a good foundation for the future improvement.

For sketch-based image retrieval (SBIR), we propose a generative adversarial network trained on a large number of sketches and their corresponding real images. To imitate human search process, we attempt to match candidate images with theimaginary image in user single s mind instead of the sketch query, i.e., not only the shape information of sketches but their possible content information are considered in SBIR. Specifically, a conditional generative adversarial network (cGAN) is employed to enrich the content information of sketches and recover the imaginary images, and two VGG-based encoders, which work on real and imaginary images respectively, are used to constrain their perceptual consistency from the view of feature representations. During SBIR, we first generate an imaginary image from a given sketch via cGAN, and then take the output of the learned encoder for imaginary images as the feature of the query sketch. Finally, we build an interactive SBIR system that shows encouraging performance.