202111 bcwh

Over the past few years, a variety of membership inference attacks against deep learning models have emerged, raising significant privacy concerns. These attacks can easily infer whether a sample exists in the training set of the target model with little adversary knowledge, and the inference accuracy is often much higher than random guessing, which causes serious privacy leakage. To this end, defenses against membership inference attacks have attracted great interest. However, the current available defense methods such as regularization, differential privacy, and knowledge distillation are unable to balance the trade-off between privacy and utility well. In this paper, we combine knowledge distillation and generative adversarial networks to propose a novel training framework that can effectively defend against membership inference attacks, called KD-GAN. Extensive experiments show that our method implements an attack success rate of nearly 0.5 (random guesses) which can successfully defend against membership inference attacks without causing significant damage to model utility, and consistently outperforming other defense methods in the balance of privacy and utility.

The vulnerabilities in cryptographic currencies facilitate the adversarial attacks. Therefore, the attackers have incentives to increase their rewards by strategic behaviors. Block withholding attacks (BWH) are such behaviors that attackers withhold blocks in the target pools to subvert the blockchain ecosystem. Furthermore, BWH attacks may dwarf the countermeasures by combining with selfish mining attacks or other strategic behaviors, for example, fork after withholding (FAW) attacks and power adaptive withholding (PAW) attacks. That is, the attackers may be intelligent enough such that they can dynamically gear their behaviors to optimal attacking strategies. In this paper, we propose mixed-BWH attacks with respect to intelligent attackers, who leverage reinforcement learning to pin down optimal strategic behaviors to maximize their rewards. More specifically, the intelligent attackers strategically toggle among BWH, FAW, and PAW attacks. Their main target is to fine-tune the optimal behaviors, which incur maximal rewards. The attackers pinpoint the optimal attacking actions with reinforcement learning, which is formalized into a Markov decision process. The simulation results show that the rewards of the mixed strategy are much higher than that of honest strategy for the attackers. Therefore, the attackers have enough incentives to adopt the mixed strategy.

The application of machine learning in the security analysis of authentication and key agreement protocol was first launched by Ma et al. in 2018. Although they received remarkable results with an accuracy of 72% for the first time, their analysis is limited to replay attack and key confirmation attack. In addition, their suggested framework is based on a multiclassification problem in which every protocol or dataset instance is either secure or prone to a security attack such as replay attack, key confirmation, or other attacks. In this paper, we show that multiclassification is not an appropriate framework for such analysis, since authentication protocols may suffer different attacks simultaneously. Furthermore, we consider more security properties and attacks to analyze protocols against. These properties include strong authentication and Unknown Key Share (UKS) attack, key freshness, key authentication, and password guessing attack. In addition, we propose a much more efficient dataset construction model using a tenth number of features, which improves the solving speed to a large extent. The results indicate that our proposed model outperforms the previous models by at least 10–20 percent in all of the machine learning solving algorithms such that upper-bound performance reaches an accuracy of over 80% in the analysis of all security properties and attacks. Despite the previous models, the classification accuracy of our proposed dataset construction model rises in a rational manner along with the increase of the dataset size.

There are thousands of malicious applications that invade Google Play Store every day and seem to be legal applications. These malicious applications have the ability to link the malware referred to as Dresscode created for network hacking as well as scrolling information. Since Android smartphones are indispensable, there should be an efficient and also unusual protection. Therefore, Android smartphones usually continue to be safeguarded from novel malware. In this paper, we propose RoughDroid, a floppy analysis technique that can discover Android malware applications directly on the smartphone. RoughDroid is based on seven feature sets ( FS1,FS2,…,FS7 ) from the XML manifest file of an Android application, plus three feature sets ( FS8,FS9, and FS10 ) from the Dex file. Those feature sets pass through the Rough Set algorithm to elastically classify the Android application as either benign or malicious. The experimental results mainly consider 20 most common malware families, plus three new malware families (Grabos, TrojanDropper.Agent.BKY, and AsiaHitGroup) that invade Google Play Store at 2017. According to the experimental results, RoughDroid has 95.6% detection performance for the malware families at 1% false-positive rate. Finally, RoughDroid is a lightweight approach for straightly examining downloaded applications on the smartphone.

The fast growth of Internet‐of‐Things (IoT) strategies has actually presented the generation of huge quantities of information. There should exist a method to conveniently gather, save, refine, and also provide such information. On the other hand, IoT data is sensitive and private information; it must not be available to potential attackers. We propose a robust scheme to guarantee both secure IoT data storage and retrieval from the untrusted cloud servers. The proposed scheme is based on Private Information Retrieval (PIR). It stores the data onto different servers and retrieves the requested data slice without disclosing its identity. In our scheme, the information is encrypted before sending to the cloud servers. It is also divided into slices of a specific size class. The experimental analysis on many different configurations supported efficiency and the efficacy of the proposed scheme, which demonstrated compatibility and exceptional performance.