Search

Can we leverage high-resolution information without the unsustainable quadratic complexity to input scale? We propose Traversal Network (TNet), a novel multi-scale hard-attention architecture, which traverses image scale-space in a top-down fashion, visiting only the most informative image regions along the way. TNet offers an adjustable trade-off between accuracy and complexity, by changing the number of attended image locations. We compare our model against hard-attention baselines on ImageNet, achieving higher accuracy with less resources (FLOPs, processing time and memory). We further test our model on fMoW dataset, where we process satellite images of size up to $896 \times 896$ px, getting up to $2.5$x faster processing compared to baselines operating on the same resolution, while achieving higher accuracy as well. TNet is modular, meaning that most classification models could be adopted as its backbone for feature extraction, making the reported performance gains orthogonal to benefits offered by existing optimized deep models. Finally, hard-attention guarantees a degree of interpretability to our model's predictions, without any extra cost beyond inference. Code is available at $\href{https://github.com/Tpap/TNet}{github.com/Tpap/TNet}$.

In this work, we examine cognitive radio networks, where secondary users may act as relays for messages sent by the primary user, hence offering performance improvement of primary transmissions, while at the same time obtaining more transmission opportunities for their own data. In particular, assuming the broadcast packet erasure model with feedback, we investigate the capacity of the fundamental cooperative cognitive radio network which consists of one primary and one secondary transmitter-receiver pairs. The primary transmitter is the owner of the channel and as such, we intend to keep its operations simple and to avoid increasing its storage requirements. Specifically, the primary transmitter does not receive data sent by the secondary transmitter and does not perform any coding operations. On the other hand, the secondary transmitter can overhear primary transmissions and is allowed to perform any coding operations. We develop an outer bound to the capacity of the fundamental cooperative cognitive radio network under consideration. Then, we propose a coding-scheduling algorithm suitable for this type of networks, which involves only XOR network coding operations. The complexity of the scheduling decisions of the proposed algorithm depends on the channel statistical parameters and three cases, depending on the relations between channel erasure probabilities, are distinguished. For the first two cases the rate region of the proposed algorithm coincides with the developed capacity outer bound, hence the algorithm is capacity achieving. For the third case, the rate region of the proposed algorithm is not identical to the outer bound; however, numerical results show that it is fairly close to the derived outer bound for a wide range of the statistical parameters of the system.

In this paper, we investigate transmission techniques for a fundamental cooperative cognitive radio network, i.e., a radio system where a Secondary user may act as relay for messages sent by the Primary user, hence offering performance improvement of Primary user transmissions, while at the same time obtaining more transmission opportunities for its own transmissions. Specifically, we examine the possibility of improving the overall system performance by employing network coding techniques. The objective is to achieve this while affecting Primary user transmissions only positively, namely: 1) avoid network coding operations at the Primary transmitter in order avoid increase of its complexity and storage requirements, 2) keep the order of packets received by the Primary receiver the same as in the non cooperative case and 3) induce packet service times that are stochastically smaller than packet service times induced in the non-cooperative case. A network coding algorithm is investigated in terms of achieved throughput region and it is shown to enlarge Secondary user throughput as compared to the case where the Secondary transmitter acts as a simple relay, while leaving the Primary user stability region unaffected. A notable feature of this algorithm is that it operates without knowledge of channel and packet arrival rate statistics. We also investigate a second network coding algorithm which increases the throughput region of the system; however, the latter algorithm requires knowledge of channel and packet arrival rate statistics.

We address the problem of shoulder-surfing attacks on authentication schemes by proposing IllusionPIN (IPIN), a PIN-based authentication method that operates on touchscreen devices. IPIN uses the technique of hybrid images to blend two keypads with different digit orderings in such a way, that the user who is close to the device is seeing one keypad to enter her PIN, while the attacker who is looking at the device from a bigger distance is seeing only the other keypad. The user's keypad is shuffled in every authentication attempt since the attacker may memorize the spatial arrangement of the pressed digits. To reason about the security of IllusionPIN, we developed an algorithm which is based on human visual perception and estimates the minimum distance from which an observer is unable to interpret the keypad of the user. We tested our estimations with 84 simulated shoulder-surfing attacks from 21 different people. None of the attacks was successful against our estimations. In addition, we estimated the minimum distance from which a camera is unable to capture the visual information from the keypad of the user. Based on our analysis, it seems practically almost impossible for a surveillance camera to capture the PIN of a smartphone user when IPIN is in use., Comment: 15 pages, 10 figures, IEEE Transactions on Information Forensics and Security, 2017

We design polynomial time schemes for secure message transmission over arbitrary networks, in the presence of an eavesdropper, and where each edge corresponds to an erasure channel with public feedback. Our schemes are described through linear programming (LP) formulations, that explicitly select (possibly different) sets of paths for key-generation and message sending. Although our LPs are not always capacity-achieving, they outperform the best known alternatives in the literature, and extend to incorporate several interesting scenaria.