“Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks”の学習報告

ABSTRACT Cyberattacks are becoming more sophisticated and ubiquitous. Cybercriminals are inevitably adopting Artificial Intelligence (AI) techniques to evade the cyberspace and cause greater damages without being noticed. Researchers in cybersecurity domain have not researched the concept behind AI-powered cyberattacks enough to understand the level of sophistication this type of attack possesses. This paper aims to investigate the emerging threat of AI-powered cyberattacks and provide insights into malicious used of AI in cyberattacks. The study was performed through a three-step process by selecting only articles based on quality, exclusion, and inclusion criteria that focus on AI-driven cyberattacks. Searches in ACM, arXiv Blackhat, Scopus, Springer, MDPI, IEEE Xplore and other sources were executed to retrieve relevant articles. Out of the 936 papers that met our search criteria, a total of 46 articles were finally selected for this study. The result shows that 56% of the AI-Driven cyberattack technique identified was demonstrated in the access and penetration phase, 12% was demonstrated in exploitation, and command and control phase, respectively; 11% was demonstrated in the reconnaissance phase; 9% was demonstrated in the delivery phase of the cybersecurity kill chain. The findings in this study shows that existing cyber defence infrastructures will become inadequate to address the increasing speed, and complex decision logic of AI-driven attacks. Hence, organizations need to invest in AI cybersecurity infrastructures to combat these emerging threats.

The Emerging Threat of Ai-driven Cyber Attacks: A Review

Make complex CAPTCHAs simple: A fast text captcha solver based on a small number of samples

Imbalance domain adaptation network with adversarial learning for fault diagnosis of rolling bearing

SybilHunter: Hybrid graph-based sybil detection by aggregating user behaviors

As a widely deployed security scheme, text-based CAPTCHAs have become more and more difficult to resist machine learning-based attacks. So far, many researchers have conducted attacking research on text-based CAPTCHAs deployed by different companies (such as Microsoft, Amazon, and Apple) and achieved certain results.However, most of these attacks have some shortcomings, such as poor portability of attack methods, requiring a series of data preprocessing steps, and relying on large amounts of labeled CAPTCHAs. In this paper, we propose an efficient and simple end-to-end attack method based on cycle-consistent generative adversarial networks. Compared with previous studies, our method greatly reduces the cost of data labeling. In addition, this method has high portability. It can attack common text-based CAPTCHA schemes only by modifying a few configuration parameters, which makes the attack easier. Firstly, we train CAPTCHA synthesizers based on the cycle-GAN to generate some fake samples. Basic recognizers based on the convolutional recurrent neural network are trained with the fake data. Subsequently, an active transfer learning method is employed to optimize the basic recognizer utilizing tiny amounts of labeled real-world CAPTCHA samples. Our approach efficiently cracked the CAPTCHA schemes deployed by 10 popular websites, indicating that our attack is likely very general. Additionally, we analyzed the current most popular anti-recognition mechanisms. The results show that the combination of more anti-recognition mechanisms can improve the security of CAPTCHA, but the improvement is limited. Conversely, generating more complex CAPTCHAs may cost more resources and reduce the availability of CAPTCHAs.

An End-to-End Attack on Text-based CAPTCHAs Based on Cycle-Consistent Generative Adversarial Network

End-to-end attack on text-based CAPTCHAs based on cycle-consistent generative adversarial network

Nowadays, millions of people use Online Social Networks (OSNs) like Twitter, Facebook and Sina Microblog, to express opinions on current events. The widespread use of these OSNs has also led to the emergence of social bots. What is more, the existence of social bots is so powerful that some of them can turn into influential users. In this paper, we studied the automated construction technology and infiltration strategies of social bots in Sina Microblog, aiming at building friendly and influential social bots to resist malicious interpretations. Firstly, we studied the critical technology of Sina Microblog data collection, which indicates that the defense mechanism of that is vulnerable. Then, we constructed 96 social bots in Sina Microblog and researched the influence of different infiltration strategies, like different attribute settings and various types of interactions. Finally, our social bots gained 5546 followers in the 42-day infiltration period with a 100% survival rate. The results show that the infiltration strategies we proposed are effective and can help social bots escape detection of Sina Microblog defense mechanism as well. The study in this paper sounds an alarm for Sina Microblog defense mechanism and provides a valuable reference for social bots detection.

/pdf/exploring-the-construction-and-infiltration-strategies-of-5joo8vitm6.pdf

Exploring the construction and infiltration strategies of social bots in sina microblog.

The invention discloses a microblog data acquisition method based on multi-strategy fusion, which comprises the following steps: firstly, simulating login, and acquiring Cookie successfully logged in;storing the obtained Cookie in a Cookie queue, and obtaining an initial task; then crawling a user attention list and user data by using a multi-account load balancing strategy; extracting a user ID,generating an attention relationship and a user information to-be-crawled queue, crawling a user attention list and user data, and generating a user microblog to-be-crawled queue; constructing a visitor Cookie, using an IP agent pool to accelerate crawling of microblog content, and storing information into a database; extracting the microblog ID, and generating a to-be-crawled queue of the comment information; crawling microblog comment information, and storing the information into a database. According to the method, the concurrent request number suitable for the current network environmentand the Cookie queue length is found through an adaptive algorithm, and balance is obtained between the collection speed and the account security. Meanwhile, a high-availability proxy IP module is realized to accelerate data acquisition, and basic data support is provided for network public opinion analysis.

Peiming Wang

Papers

An End-to-End Attack on Text-based CAPTCHAs Based on Cycle-Consistent Generative Adversarial Network

End-to-end attack on text-based CAPTCHAs based on cycle-consistent generative adversarial network

Exploring the construction and infiltration strategies of social bots in sina microblog.

Microblog data acquisition method based on multi-strategy fusion