Biblio

Model compression is one of the most preferred techniques for efficiently deploying deep neural networks (DNNs) on resource- constrained Internet of Things (IoT) platforms. However, the simply compressed model is often vulnerable to adversarial attacks, leading to a conflict between robustness and efficiency, especially for IoT devices exposed to complex real-world scenarios. We, for the first time, address this problem by developing a novel framework dubbed Magical-Decomposition to simultaneously enhance both robustness and efficiency for hardware. By leveraging a hardware-friendly model compression method called singular value decomposition, the defending algorithm can be supported by most of the existing DNN hardware accelerators. To step further, by using a recently developed DNN interpretation tool, the underlying scheme of how the adversarial accuracy can be increased in the compressed model is highlighted clearly. Ablation studies and extensive experiments under various attacks/models/datasets consistently validate the effectiveness and scalability of the proposed framework.

With the rapid development of multimedia and short video, there is a growing concern for video copyright protection. Some work has been proposed to add some copyright or fingerprint information to the video to trace the source of the video when it is stolen and protect video copyright. This paper proposes a video watermarking method based on a deep neural network and curriculum learning for watermarking of sliced videos. The first frame of the segmented video is perturbed by an encoder network, which is invisible and can be distinguished by the decoder network. Our model is trained and tested on an online educational video dataset consisting of 2000 different video clips. Experimental results show that our method can successfully discriminate most watermarked and non-watermarked videos with low visual disturbance, which can be achieved even under a relatively high video compression rate(H.264 video compress with CRF 32).

True random number generators (TRNG) are widely used to generate encryption keys in information security systems [1]–[2]. In TRNG, entropy source is a critical module who provides the source of randomness of output bit stream. The unavoidable electrical noise in circuit becomes an ideal entropy source due to its unpredictability. Among the methods of capturing electrical noise, ring oscillator-based entropy source makes the TRNG most robust to deterministic noise and 1/f noise which means the strongest anti-interference capability, so it is simple in structure and easy to integrate [3]. Thus, great research attention has focused on ring oscillator-based TRNGs [3] –[7]. In [4], a high-speed TRNG with 100Mbps output bit rate was proposed, but it took up too much power and area. A TRNG based on tetrahedral ring oscillator was proposed in [5]. Its power consumption was very low but the output bit rate was also very low. A ring oscillator-based TRNG with low output bit rate but high power was proposed in [7]. In a word, none of the above architectures achieve an appropriate compromise between bit rate and power consumption. This work presents a new feedback architecture of TRNG based on tetrahedral ring oscillator. The output random bit stream generates a relative random control voltage that acts on the transmission gates in oscillator through a feedback loop, thus increasing phase jitter of the oscillator and improving output bit rate. Furthermore, an XOR chain-based post-processing unit is added to eliminate the statistical deviations and correlations between raw bits.

Encryption schemes for network security usually require a key distribution center to share or distribute the secret keys, which is difficult to deploy in wireless networks without fixed infrastructure. A novel key generation scheme based on the physical layer can generate a shared key between a pair of correlated parties by sharing random sources. The existing physical layer key generation scheme is based on the half-duplex mode with time division duplex (TDD) mode, which makes it impossible for the correlated communication parties to detect the channel simultaneously in order to improve the channel coherence. In this paper, we propose a full-duplex physical layer key generation scheme, which allows each legal communication nodes to transmit and receive signals at the same time, in order to reduce channel probing time and increase channel coherence performance. The simulation experiments show that the proposed scheme can much outperform some typical existing schemes in terms of the key performance evaluation indicators, key disagreement rate, key generation rate, entropy of the scheme improved, and the randomness of generated keys passed the National Institute of Standards and Technology (NIST) test.