Deep Neural Network (DNN) watermarking techniques are increasingly being used to protect the intellectual property of DNN models. Basically, DNN watermarking is a technique to insert side information into the DNN model without significantly degrading the performance of its original task. A pruning attack is a threat to DNN watermarking, wherein the less important neurons in the model are pruned to make it faster and more compact. As a result, removing the watermark from the DNN model is possible. This study investigates a channel coding approach to protect DNN watermarking against pruning attacks. The channel model differs completely from conventional models involving digital images. Determining the suitable encoding methods for DNN watermarking remains an open problem. Herein, we presented a novel encoding approach using constant weight codes to protect the DNN watermarking against pruning attacks. The experimental results confirmed that the robustness against pruning attacks could be controlled by carefully setting two thresholds for binary symbols in the codeword.

No abstract

Recently in many rivers, vegetation covers wide parts of floodplain of fluvial-fan segment of particularly with a dam on the upstream. Rich vegetation is disadvantageous from the viewpoint of safety against flood but also it is not necessary good for ecosystem of fluvial-fan rivers. When the changes of river landscape of the Tedori river (with dam) and the Abe river (without dam) are compared with each other by aerial-photo analysis, the growth of vegetation after dam construction is remarkable in the Tedori river. When the records of floods of the two rivers are compared with each other, it is clarified how the major floods were controlled by a dam in the Tedori river. The previous studies by us show that repetition of medium-size floods and low-stage waters promotes vegetated area to expand. On the other hand, major floods sometimes destroy the vegetation. Usually the growth and destruction are in balance. Thus the obvious growth of vegetation in the Tedori river probably implies the decrease of chances for vegetation to be destroyed by major floods. Then, models to simulate destruction of vegetation by flood and growth of individual trees have been proposed to explain the difference of landscape changes in the Tedori and the Abe river.

A previously proposed optimal detector for biasbased fingerprinting codes such as Tardos and Nuida requires two kinds of important information: the number of colluders and the collusion strategy used to generate the pirated codeword. An estimator has now been derived for these two parameters. The bias in the pirated codeword is measured by observing the number of zeros and ones and compared with possible bias patterns calculated using information about the collusion strategy and number of colluders. Computer simulation demonstrated that the collusion strategy and number of colluders can be estimated with high probability and that the traceability of a detector using the proposed estimator is extremely close to being optimal.

No abstract