To meet the requirements of real-world applications, it is essential to control generations of la... more To meet the requirements of real-world applications, it is essential to control generations of large language models (LLMs). Prior research has tried to introduce reinforcement learning (RL) into controllable text generation while most existing methods suffer from overfitting issues (finetuning-based methods) or semantic collapse (post-processing methods). However, current RL methods are generally guided by coarse-grained (sentence/paragraph-level) feedback, which may lead to suboptimal performance owing to semantic twists or progressions within sentences. To tackle that, we propose a novel reinforcement learning algorithm named TOLE which formulates TOken-LEvel rewards for controllable text generation, and employs a "first-quantize-then-noise" paradigm to enhance the robustness of the RL algorithm. Furthermore, TOLE can be flexibly extended to multiple constraints with little computational expense. Experimental results show that our algorithm can achieve superior performance on both single-attribute and multi-attribute control tasks. We have released our codes at https://github.com/WindyLee0822/CTG.
By exploiting multipath fading channels as a source of common randomness, physical layer (PHY) ba... more By exploiting multipath fading channels as a source of common randomness, physical layer (PHY) based key generation protocols allow two terminals with correlated observations to generate secret keys with information-theoretical security. The state of the art, however, still suffers from major limitations, e.g., low key generation rate, lower entropy of key bits and a high reliance on node mobility. In this paper, a novel cooperative key generation protocol is developed to facilitate high-rate key generation in narrowband fading channels, where two keying nodes extract the phase randomness of the fading channel with the aid of relay node(s). For the first time, we explicitly consider the effect of estimation methods on the extraction of secret key bits from the underlying fading channels and focus on a popular statistical method--maximum likelihood estimation (MLE). The performance of the cooperative key generation scheme is extensively evaluated theoretically. We successfully establ...
By exploiting multipath fading channels as a source of common randomness, physical layer (PHY) ba... more By exploiting multipath fading channels as a source of common randomness, physical layer (PHY) based key generation protocols allow two terminals with correlated observations to generate secret keys with information-theoretical security. The state of the art, however, still suffers from major limitations, e.g., low key generation rate, lower entropy of key bits and a high reliance on node mobility. In this paper, a novel cooperative key generation protocol is developed to facilitate high-rate key generation in narrowband fading channels, where two keying nodes extract the phase randomness of the fading channel with the aid of relay node(s). For the first time, we explicitly consider the effect of estimation methods on the extraction of secret key bits from the underlying fading channels and focus on a popular statistical method-maximum likelihood estimation (MLE). The performance of the cooperative key generation scheme is extensively evaluated theoretically. We successfully establish both a theoretical upper bound on the maximum secret key rate from mutual information of correlated random sources and a more practical upper bound from Cramer-Rao bound (CRB) in estimation theory. Numerical examples and simulation studies are also presented to demonstrate the performance of the cooperative key generation system. The results show that the key rate can be improved by a couple of orders of magnitude compared to the existing approaches.
This paper presents a novel approach to train GANs for discrete sequence generation without resor... more This paper presents a novel approach to train GANs for discrete sequence generation without resorting to an explicit neural network as the discriminator. We show that when an alternative mini-max optimization procedure is performed for the value function where a closed form solution for the discriminator exists in the maximization step, it is equivalent to directly optimizing the Jenson-Shannon divergence (JSD) between the generator’s distribution and the empirical distribution over the training data without sampling from the generator, thus optimizing the JSD becomes computationally tractable to train the generator that generates sequences of discrete data. Extensive experiments on synthetic data and real-world tasks demonstrate significant improvements over existing methods to train GANs that generate discrete sequences.
2015 IEEE Global Communications Conference (GLOBECOM), 2014
Advancements in big data analysis offer costeffective opportunities to improve decision-making in... more Advancements in big data analysis offer costeffective opportunities to improve decision-making in numerous areas such as health care, economic productivity, crime, and resource management. Nowadays, data holders are tending to sharing their data for better outcomes from their aggregated data. However, the current tools and technologies developed to manage big data are often not designed to incorporate adequate security or privacy measures during data sharing. In this paper, we consider a scenario where multiple data holders intend to find predictive models from their joint data without revealing their own data to each other. Data locality property is used as an alternative to multi-party computation (SMC) techniques. Specifically, we distribute the centralized learning task to each data holder as local learning tasks in a way that local learning is only related to local data. Along with that, we propose an efficient and secure protocol to reassemble local results to get the final result. Correctness of our scheme is proved theoretically and numerically. Security analysis is conducted from the aspect of information theory.
2015 IEEE 35th International Conference on Distributed Computing Systems, 2015
Machine learning has played an increasing important role in big data systems due to its capabilit... more Machine learning has played an increasing important role in big data systems due to its capability of efficiently discovering valuable knowledge and hidden information. Often times big data such as healthcare systems or financial systems may involve with multiple organizations who may have different privacy policy, and may not explicitly share their data publicly while joint data processing may be a must. Thus, how to share big data among distributed data processing entities while mitigating privacy concerns becomes a challenging problem. Traditional methods rely on cryptographic tools and/or randomization to preserve privacy. Unfortunately, this alone may be inadequate for the emerging big data systems because they are mainly designed for traditional small-scale data sets. In this paper, we propose a novel framework to achieve privacy-preserving machine learning where the training data are distributed and each shared data portion is of large volume. Specifically, we utilize the data locality property of Apache Hadoop architecture and only a limited number of cryptographic operations at the Reduce() procedures to achieve privacy-preservation. We show that the proposed scheme is secure in the semi-honest model and use extensive simulations to demonstrate its scalability and correctness.
IEEE Transactions on Dependable and Secure Computing, 2017
demonstrate that our framework is better than other conceivable methodologies as far as acknowled... more demonstrate that our framework is better than other conceivable methodologies as far as acknowledgment proportion and effectiveness. Our component is actualized as a proof of idea Android application on Facebook's platform.The control law dissemination is caused by the special join process, in which the likelihood of a client An interfacing with a client B is relative to the quantity of B's current associations. demonstrate the previews of the contact system and fan arrange in YA, separately. We see a few hubs don't have either fans or contacts, while a couple of hubs have a substantial degree.
Photo sharing is an attractive feature which popularizes Online Social Networks (OSNs). Unfortuna... more Photo sharing is an attractive feature which popularizes Online Social Networks (OSNs). Unfortunately, it may leak users' privacy if they are allowed to post, comment, and tag a photo freely. In this paper, we attempt to address this issue and study the scenario when a user shares a photo containing individuals other than himself/herself (termed co-photo for short). To prevent possible leakage of a photo privacy, we design a mechanism to enable each individual in a photo be aware of the posting activity and participate in the decision making on the photo posting. For this purpose, we need an efficient facial recognition (FR) system that can recognize everyone in the photo. However, more demanding privacy setting may limit the number of the photos publicly available to train the FR system. To deal with this dilemma, our mechanism attempts to utilize users' private photos to design a personalized FR system specifically trained to differentiate possible photo co-owners without leaking his/her privacy. We have also developed a distributed consensusbased method to not only reduce the computational complexity, but also preserve the privacy during the training. We show that our system is superior to other possible approaches in terms of recognition ratio and efficiency. Our mechanism is implemented as an Android application on Facebook's platform.
2011 - MILCOM 2011 Military Communications Conference, 2011
Recently, there has been great research interest in utilizing the phase response of multipath fad... more Recently, there has been great research interest in utilizing the phase response of multipath fading channels as sources of common randomness to achieve information-theoretic key establishment. Previous work along this direction assume a common time reference is shared among the keying nodes such that each node can obtain absolute phase estimations. However, such assumption is impractical in some scenarios where common time reference is not available. Without the common time reference, each keying node in the system has to estimate the channel phase based on its own independent and imperfect local clock, resulting in a phase offset between its local estimates and the absolute estimates. In this paper, we study the effect of clock deviations among keying nodes and propose a novel key generation protocol with asynchronous clocks among the keying nodes. Our contributions include: 1) Mitigate the reliance on common time reference to get a desired key agreement probability; 2) Improve the key generation rate by utilizing additional randomness provided by the relay nodes. Numerical and simulation results show that a desired estimation accuracy and key generation rate could be achieved with the help of a rely node.
IEEE Journal on Selected Areas in Communications, 2012
By exploiting multipath fading channels as a source of common randomness, physical layer (PHY) ba... more By exploiting multipath fading channels as a source of common randomness, physical layer (PHY) based key generation protocols allow two terminals with correlated observations to generate secret keys with information-theoretical security. The state of the art, however, still suffers from major limitations, e.g., low key generation rate, lower entropy of key bits and a high reliance on node mobility. In this paper, a novel cooperative key generation protocol is developed to facilitate high-rate key generation in narrowband fading channels, where two keying nodes extract the phase randomness of the fading channel with the aid of relay node(s). For the first time, we explicitly consider the effect of estimation methods on the extraction of secret key bits from the underlying fading channels and focus on a popular statistical method-maximum likelihood estimation (MLE). The performance of the cooperative key generation scheme is extensively evaluated theoretically. We successfully establish both a theoretical upper bound on the maximum secret key rate from mutual information of correlated random sources and a more practical upper bound from Cramer-Rao bound (CRB) in estimation theory. Numerical examples and simulation studies are also presented to demonstrate the performance of the cooperative key generation system. The results show that the key rate can be improved by a couple of orders of magnitude compared to the existing approaches.
Jamming-resistant communication without preshared secrets has received extensive attention recent... more Jamming-resistant communication without preshared secrets has received extensive attention recently and is commonly tackled by utilizing the technique of uncoordinated frequency hopping (UFH). However, existing approaches exhibit significant performance constraints due to the use of UFH at both the sender and the receiver sides. To improve the state of the art, in this paper we aim to significantly improve the performance of the anti-jamming system in the presence of a power-limited jammer. Specifically, we for the first time jointly consider UFH and power control and pose these two techniques into a uniform framework. The proposed approach utilizes online learning theory to determine both the hopping channels and the transmitting powers based on the history of channel status. By dividing the transmission power into multiple levels, the sender with a limited power budget is able to choose both the sending channels and the corresponding transmission power. The sender keeps refining its knowledge of channel status to improve future channel selection and power allocation based on the feedback information from the receiver. We analytically show that, in presence of a power-limited jammer, the average transmission delay of our system is bounded by 2A 2T |Cs| ln |Ns| with high probability. Extensive simulations are conducted to demonstrate the effectiveness of our scheme against various jamming attacks.
To meet the requirements of real-world applications, it is essential to control generations of la... more To meet the requirements of real-world applications, it is essential to control generations of large language models (LLMs). Prior research has tried to introduce reinforcement learning (RL) into controllable text generation while most existing methods suffer from overfitting issues (finetuning-based methods) or semantic collapse (post-processing methods). However, current RL methods are generally guided by coarse-grained (sentence/paragraph-level) feedback, which may lead to suboptimal performance owing to semantic twists or progressions within sentences. To tackle that, we propose a novel reinforcement learning algorithm named TOLE which formulates TOken-LEvel rewards for controllable text generation, and employs a "first-quantize-then-noise" paradigm to enhance the robustness of the RL algorithm. Furthermore, TOLE can be flexibly extended to multiple constraints with little computational expense. Experimental results show that our algorithm can achieve superior performance on both single-attribute and multi-attribute control tasks. We have released our codes at https://github.com/WindyLee0822/CTG.
By exploiting multipath fading channels as a source of common randomness, physical layer (PHY) ba... more By exploiting multipath fading channels as a source of common randomness, physical layer (PHY) based key generation protocols allow two terminals with correlated observations to generate secret keys with information-theoretical security. The state of the art, however, still suffers from major limitations, e.g., low key generation rate, lower entropy of key bits and a high reliance on node mobility. In this paper, a novel cooperative key generation protocol is developed to facilitate high-rate key generation in narrowband fading channels, where two keying nodes extract the phase randomness of the fading channel with the aid of relay node(s). For the first time, we explicitly consider the effect of estimation methods on the extraction of secret key bits from the underlying fading channels and focus on a popular statistical method--maximum likelihood estimation (MLE). The performance of the cooperative key generation scheme is extensively evaluated theoretically. We successfully establ...
By exploiting multipath fading channels as a source of common randomness, physical layer (PHY) ba... more By exploiting multipath fading channels as a source of common randomness, physical layer (PHY) based key generation protocols allow two terminals with correlated observations to generate secret keys with information-theoretical security. The state of the art, however, still suffers from major limitations, e.g., low key generation rate, lower entropy of key bits and a high reliance on node mobility. In this paper, a novel cooperative key generation protocol is developed to facilitate high-rate key generation in narrowband fading channels, where two keying nodes extract the phase randomness of the fading channel with the aid of relay node(s). For the first time, we explicitly consider the effect of estimation methods on the extraction of secret key bits from the underlying fading channels and focus on a popular statistical method-maximum likelihood estimation (MLE). The performance of the cooperative key generation scheme is extensively evaluated theoretically. We successfully establish both a theoretical upper bound on the maximum secret key rate from mutual information of correlated random sources and a more practical upper bound from Cramer-Rao bound (CRB) in estimation theory. Numerical examples and simulation studies are also presented to demonstrate the performance of the cooperative key generation system. The results show that the key rate can be improved by a couple of orders of magnitude compared to the existing approaches.
This paper presents a novel approach to train GANs for discrete sequence generation without resor... more This paper presents a novel approach to train GANs for discrete sequence generation without resorting to an explicit neural network as the discriminator. We show that when an alternative mini-max optimization procedure is performed for the value function where a closed form solution for the discriminator exists in the maximization step, it is equivalent to directly optimizing the Jenson-Shannon divergence (JSD) between the generator’s distribution and the empirical distribution over the training data without sampling from the generator, thus optimizing the JSD becomes computationally tractable to train the generator that generates sequences of discrete data. Extensive experiments on synthetic data and real-world tasks demonstrate significant improvements over existing methods to train GANs that generate discrete sequences.
2015 IEEE Global Communications Conference (GLOBECOM), 2014
Advancements in big data analysis offer costeffective opportunities to improve decision-making in... more Advancements in big data analysis offer costeffective opportunities to improve decision-making in numerous areas such as health care, economic productivity, crime, and resource management. Nowadays, data holders are tending to sharing their data for better outcomes from their aggregated data. However, the current tools and technologies developed to manage big data are often not designed to incorporate adequate security or privacy measures during data sharing. In this paper, we consider a scenario where multiple data holders intend to find predictive models from their joint data without revealing their own data to each other. Data locality property is used as an alternative to multi-party computation (SMC) techniques. Specifically, we distribute the centralized learning task to each data holder as local learning tasks in a way that local learning is only related to local data. Along with that, we propose an efficient and secure protocol to reassemble local results to get the final result. Correctness of our scheme is proved theoretically and numerically. Security analysis is conducted from the aspect of information theory.
2015 IEEE 35th International Conference on Distributed Computing Systems, 2015
Machine learning has played an increasing important role in big data systems due to its capabilit... more Machine learning has played an increasing important role in big data systems due to its capability of efficiently discovering valuable knowledge and hidden information. Often times big data such as healthcare systems or financial systems may involve with multiple organizations who may have different privacy policy, and may not explicitly share their data publicly while joint data processing may be a must. Thus, how to share big data among distributed data processing entities while mitigating privacy concerns becomes a challenging problem. Traditional methods rely on cryptographic tools and/or randomization to preserve privacy. Unfortunately, this alone may be inadequate for the emerging big data systems because they are mainly designed for traditional small-scale data sets. In this paper, we propose a novel framework to achieve privacy-preserving machine learning where the training data are distributed and each shared data portion is of large volume. Specifically, we utilize the data locality property of Apache Hadoop architecture and only a limited number of cryptographic operations at the Reduce() procedures to achieve privacy-preservation. We show that the proposed scheme is secure in the semi-honest model and use extensive simulations to demonstrate its scalability and correctness.
IEEE Transactions on Dependable and Secure Computing, 2017
demonstrate that our framework is better than other conceivable methodologies as far as acknowled... more demonstrate that our framework is better than other conceivable methodologies as far as acknowledgment proportion and effectiveness. Our component is actualized as a proof of idea Android application on Facebook's platform.The control law dissemination is caused by the special join process, in which the likelihood of a client An interfacing with a client B is relative to the quantity of B's current associations. demonstrate the previews of the contact system and fan arrange in YA, separately. We see a few hubs don't have either fans or contacts, while a couple of hubs have a substantial degree.
Photo sharing is an attractive feature which popularizes Online Social Networks (OSNs). Unfortuna... more Photo sharing is an attractive feature which popularizes Online Social Networks (OSNs). Unfortunately, it may leak users' privacy if they are allowed to post, comment, and tag a photo freely. In this paper, we attempt to address this issue and study the scenario when a user shares a photo containing individuals other than himself/herself (termed co-photo for short). To prevent possible leakage of a photo privacy, we design a mechanism to enable each individual in a photo be aware of the posting activity and participate in the decision making on the photo posting. For this purpose, we need an efficient facial recognition (FR) system that can recognize everyone in the photo. However, more demanding privacy setting may limit the number of the photos publicly available to train the FR system. To deal with this dilemma, our mechanism attempts to utilize users' private photos to design a personalized FR system specifically trained to differentiate possible photo co-owners without leaking his/her privacy. We have also developed a distributed consensusbased method to not only reduce the computational complexity, but also preserve the privacy during the training. We show that our system is superior to other possible approaches in terms of recognition ratio and efficiency. Our mechanism is implemented as an Android application on Facebook's platform.
2011 - MILCOM 2011 Military Communications Conference, 2011
Recently, there has been great research interest in utilizing the phase response of multipath fad... more Recently, there has been great research interest in utilizing the phase response of multipath fading channels as sources of common randomness to achieve information-theoretic key establishment. Previous work along this direction assume a common time reference is shared among the keying nodes such that each node can obtain absolute phase estimations. However, such assumption is impractical in some scenarios where common time reference is not available. Without the common time reference, each keying node in the system has to estimate the channel phase based on its own independent and imperfect local clock, resulting in a phase offset between its local estimates and the absolute estimates. In this paper, we study the effect of clock deviations among keying nodes and propose a novel key generation protocol with asynchronous clocks among the keying nodes. Our contributions include: 1) Mitigate the reliance on common time reference to get a desired key agreement probability; 2) Improve the key generation rate by utilizing additional randomness provided by the relay nodes. Numerical and simulation results show that a desired estimation accuracy and key generation rate could be achieved with the help of a rely node.
IEEE Journal on Selected Areas in Communications, 2012
By exploiting multipath fading channels as a source of common randomness, physical layer (PHY) ba... more By exploiting multipath fading channels as a source of common randomness, physical layer (PHY) based key generation protocols allow two terminals with correlated observations to generate secret keys with information-theoretical security. The state of the art, however, still suffers from major limitations, e.g., low key generation rate, lower entropy of key bits and a high reliance on node mobility. In this paper, a novel cooperative key generation protocol is developed to facilitate high-rate key generation in narrowband fading channels, where two keying nodes extract the phase randomness of the fading channel with the aid of relay node(s). For the first time, we explicitly consider the effect of estimation methods on the extraction of secret key bits from the underlying fading channels and focus on a popular statistical method-maximum likelihood estimation (MLE). The performance of the cooperative key generation scheme is extensively evaluated theoretically. We successfully establish both a theoretical upper bound on the maximum secret key rate from mutual information of correlated random sources and a more practical upper bound from Cramer-Rao bound (CRB) in estimation theory. Numerical examples and simulation studies are also presented to demonstrate the performance of the cooperative key generation system. The results show that the key rate can be improved by a couple of orders of magnitude compared to the existing approaches.
Jamming-resistant communication without preshared secrets has received extensive attention recent... more Jamming-resistant communication without preshared secrets has received extensive attention recently and is commonly tackled by utilizing the technique of uncoordinated frequency hopping (UFH). However, existing approaches exhibit significant performance constraints due to the use of UFH at both the sender and the receiver sides. To improve the state of the art, in this paper we aim to significantly improve the performance of the anti-jamming system in the presence of a power-limited jammer. Specifically, we for the first time jointly consider UFH and power control and pose these two techniques into a uniform framework. The proposed approach utilizes online learning theory to determine both the hopping channels and the transmitting powers based on the history of channel status. By dividing the transmission power into multiple levels, the sender with a limited power budget is able to choose both the sending channels and the corresponding transmission power. The sender keeps refining its knowledge of channel status to improve future channel selection and power allocation based on the feedback information from the receiver. We analytically show that, in presence of a power-limited jammer, the average transmission delay of our system is bounded by 2A 2T |Cs| ln |Ns| with high probability. Extensive simulations are conducted to demonstrate the effectiveness of our scheme against various jamming attacks.
Uploads
Papers by Kaihe Xu