Ilias Leontiadis's research while affiliated with Cambridge Crystallographic Data Centre and other places

Publications (55)

Article
Recently, there has been an explosive growth of mobile and embedded applications using convolutional neural networks (CNNs). To alleviate their excessive computational demands, developers have traditionally resorted to cloud offloading, inducing high infrastructure costs and a strong dependence on networking conditions. On the other end, the emerge...
Preprint
With smartphones' omnipresence in people's pockets, Machine Learning (ML) on mobile is gaining traction as devices become more powerful. With applications ranging from visual filters to voice assistants, intelligence on mobile comes in many forms and facets. However, Deep Neural Network (DNN) inference remains a compute intensive workload, with dev...
Preprint
Full-text available
The unprecedented performance of deep neural networks (DNNs) has led to large strides in various Artificial Intelligence (AI) inference tasks, such as object and speech recognition. Nevertheless, deploying such AI models across commodity devices faces significant challenges: large computational cost, multiple performance objectives, hardware hetero...
Preprint
Recently, there has been an explosive growth of mobile and embedded applications using convolutional neural networks(CNNs). To alleviate their excessive computational demands, developers have traditionally resorted to cloud offloading, inducing high infrastructure costs and a strong dependence on networking conditions. On the other end, the emergen...
Preprint
On-device machine learning is becoming a reality thanks to the availability of powerful hardware and model compression techniques. Typically, these models are pretrained on large GPU clusters and have enough parameters to generalise across a wide variety of inputs. In this work, we observe that a much smaller, personalised model can be employed to...
Preprint
Despite the soaring use of convolutional neural networks (CNNs) in mobile applications, uniformly sustaining high-performance inference on mobile has been elusive due to the excessive computational demands of modern CNNs and the increasing diversity of deployed devices. A popular alternative comprises offloading CNN processing to powerful cloud-bas...
Article
We have reached an important milestone in Automatic Speech Recognition (ASR) technology, with major industrial AI companies, such as Samsung, Google, Apple, and Amazon releasing high-quality ASR models that run completely on-device, e.g., on consumer smartphones. This is the consequence of giant strides in technological advancements: from making co...
Preprint
In late 2017, a sudden proliferation of malicious JavaScript was reported on the Web: browser-based mining exploited the CPU time of website visitors to mine the cryptocurrency Monero. Several studies measured the deployment of such code and developed defenses. However, previous work did not establish how many users were really exposed to the ident...
Preprint
Full-text available
We present DarkneTZ, a framework that uses an edge device's Trusted Execution Environment (TEE) in conjunction with model partitioning to limit the attack surface against Deep Neural Networks (DNNs). Increasingly, edge devices (smartphones and consumer IoT devices) are equipped with pre-trained DNNs for a variety of applications. This trend comes w...
Conference Paper
Remembering our day-to-day social interactions is challenging even if you aren't a blue memory challenged fish. The ability to automatically detect and remember these types of interactions is not only beneficial for individuals interested in their behavior in crowded situations, but also of interest to those who analyze crowd behavior. Currently, d...
Article
Full-text available
Video sharing platforms like YouTube are increasingly targeted by aggression and hate attacks. Prior work has shown how these attacks often take place as a result of "raids," i.e., organized efforts by ad-hoc mobs coordinating from third-party communities. Despite the increasing relevance of this phenomenon, however, online services often lack effe...
Article
Full-text available
Cyberbullying and cyberaggression are increasingly worrisome phenomena affecting people across all demographics. More than half of young social media users worldwide have been exposed to such prolonged and/or coordinated digital harassment. Victims can experience a wide range of emotions, with negative consequences such as embarrassment, depression...
Conference Paper
In recent years, we observed an increased importance of experimental and empirical work on wireless research. In this talk we will thus describe industrial testbeds and operational environments and the research that can be done on them. Specifically, we will focus on Telefonica and Samsung research providing examples on operational cellular network...
Preprint
Full-text available
Cyberbullying and cyberaggression are increasingly worrisome phenomena affecting people across all demographics. More than half of young social media users worldwide have been exposed to such prolonged and/or coordinated digital harassment. Victims can experience a wide range of emotions, with negative consequences such as embarrassment, depression...
Conference Paper
Hate speech, offensive language, sexism, racism, and other types of abusive behavior have become a common phenomenon in many online social media platforms. In recent years, such diverse abusive behaviors have been manifesting with increased frequency and levels of intensity. Despite social media's efforts to combat online abusive behaviors this pro...
Conference Paper
In recent years, advances in deep learning have resulted in unprecedented leaps in diverse tasks spanning from speech and object recognition to context awareness and health monitoring. As a result, an increasing number of AI-enabled applications are being developed targeting ubiquitous and mobile devices. While deep neural networks (DNNs) are getti...
Preprint
In recent years, advances in deep learning have resulted in unprecedented leaps in diverse tasks spanning from speech and object recognition to context awareness and health monitoring. As a result, an increasing number of AI-enabled applications are being developed targeting ubiquitous and mobile devices. While deep neural networks (DNNs) are getti...
Preprint
Full-text available
Modern deep learning approaches have achieved groundbreaking performance in modeling and classifying sequential data. Specifically, attention networks constitute the state-of-the-art paradigm for capturing long temporal dynamics. This paper examines the efficacy of this paradigm in the challenging task of emotion recognition in dyadic conversations...
Conference Paper
Full-text available
Edge computing is considered a key enabler to deploy Artificial Intelligence platforms to provide real-time applications such as AR/VR or cognitive assistance. Previous works show computing capabilities deployed very close to the user can actually reduce the end-to-end latency of such interactive applications. Nonetheless, the main performance bott...
Preprint
Full-text available
A large number of the most-subscribed YouTube channels target children of very young age. Hundreds of toddler-oriented channels on YouTube feature inoffensive, well produced, and educational videos. Unfortunately, inappropriate content that targets this demographic is also common. YouTube’s algorithmic recommendation system regrettably suggests ina...
Conference Paper
Full-text available
Botnets in online social networks are increasingly often affecting the regular flow of discussion, attacking regular users and their posts, spamming them with irrelevant or offensive content, and even manipulating the popularity of messages and accounts. Researchers and cybercriminals are involved in an arms race, and new and updated botnets design...
Conference Paper
Full-text available
Video sharing platforms like YouTube are increasingly targeted by aggression and hate attacks. Prior work has shown how these attacks often take place as a result of “raids,” i.e., organized efforts by ad-hoc mobs coordinating from third-party communities. Despite the increasing relevance of this phenomenon, however, online services often lack effe...
Conference Paper
Recent advances are driving wearables towards stand-alone devices with cellular network support (e.g. SIM-enabled Apple Watch series-3). Nonetheless, a little has been studied on SIM-enabled wearable traffic in ISP networks to gain customer insights and to understand traffic characteristics. In this paper, we characterize the network traffic of sev...
Preprint
Full-text available
Botnets in online social networks are increasingly often affecting the regular flow of discussion, attacking regular users and their posts, spamming them with irrelevant or offensive content, and even manipulating the popularity of messages and accounts. Researchers and cybercriminals are involved in an arms race, and new and updated botnets design...
Preprint
Full-text available
Remembering our day-to-day social interactions is challenging even if you aren't a blue memory challenged fish. The ability to automatically detect and remember these types of interactions is not only beneficial for individuals interested in their behavior in crowded situations, but also of interest to those who analyze crowd behavior. Currently, d...
Preprint
Full-text available
With current technology, a number of entities have access to user mobility traces at different levels of spatio-temporal granularity. At the same time, users frequently reveal their location through different means, including geo-tagged social media posts and mobile app usage. Such leaks are often bound to a pseudonym or a fake identity in an attem...
Preprint
Full-text available
Over the years, the Web has shrunk the world, allowing individuals to share viewpoints with many more people than they are able to in real life. At the same time, however, it has also enabled anti-social and toxic behavior to occur at an unprecedented scale. Video sharing platforms like YouTube receive uploads from millions of users, covering a wid...
Conference Paper
Full-text available
While developing mobile apps is becoming easier, testing and characterizing their behavior is still hard. On the one hand, the de facto testing tool, called "Monkey," scales well due to being based on random inputs, but fails to gather inputs useful in understanding things like user engagement and attention. On the other hand, gathering inputs and...
Article
Full-text available
Hate speech, offensive language, sexism, racism and other types of abusive behavior have become a common phenomenon in many online social media platforms. In recent years, such diverse abusive behaviors have been manifesting with increased frequency and levels of intensity. This is due to the openness and willingness of popular media platforms, suc...
Conference Paper
Full-text available
In recent years, offensive, abusive and hateful language, sexism, racism and other types of aggressive and cyberbullying behavior have been manifesting with increased frequency, and in many online social media platforms. In fact, past scientific work focused on studying these forms in popular media, such as Facebook and Twitter. Building on such wo...
Article
Full-text available
We investigate to what extent mobile use patterns can predict -- at the moment it is posted -- whether a notification will be clicked within the next 10 minutes. We use a data set containing the detailed mobile phone usage logs of 279 users, who over the course of 5 weeks received 446,268 notifications from a variety of apps. Besides using classica...
Conference Paper
Full-text available
As the number and the diversity of news outlets on the Web grows, so does the opportunity for "alternative" sources of information to emerge. Using large social networks like Twitter and Facebook, misleading, false, or agenda-driven information can quickly and seamlessly spread online, deceiving people or influencing their opinions. Also, the incre...
Conference Paper
Full-text available
Network performance anomalies can be defined as abnormal and significant variations in a network’s traffic levels. Being able to detect anomalies is critical for both network operators and end users. However, the accurate detection without raising false alarms can become a challenging task when there is high variance in the traffic. To address this...
Conference Paper
Full-text available
Common approaches to text categorization essentially rely either on n-gram counts or on word embeddings. This presents important difficulties in highly dynamic or quickly-interacting environments, where the appearance of new words and/or varied misspellings is the norm. A paradigmatic example of this situation is abusive online behavior, with socia...
Article
We present a practical approach for processing mobile sensor time series data for continual deep learning predictions. The approach comprises data cleaning, normalization, capping, time-based compression, and finally classification with a recurrent neural network. We demonstrate the effectiveness of the approach in a case study with 279 participant...
Conference Paper
Full-text available
Although it has been a part of the dark underbelly of the Internet since its inception, recent events have brought the discussion board site 4chan to the forefront of the world's collective mind. In particular, /pol/, 4chan's "Politically Incorrect" board has become a central figure in the outlandish 2016 Presidential election. Even though 4chan ha...
Article
Full-text available
To manage and maintain large-scale cellular networks, operators need to know which sectors underperform at any given time. For this purpose, they use the so-called hot spot score, which is the result of a combination of multiple network measurements and reflects the instantaneous overall performance of individual sectors. While operators have a goo...
Conference Paper
Full-text available
Mobile network operators collect a humongous amount of network measurements. Among those, sector Key Performance Indicators (KPIs) are used to monitor the radio access, i.e., the “last mile” of mobile networks. Thresholding mechanisms and synthetic combinations of KPIs are used to assess the network health, and rank sectors to identify the underper...
Conference Paper
Full-text available
Tracking and maintaining satisfactory QoE for video streaming services is becoming a greater challenge for mobile network operators than ever before. Downloading and watching video content on mobile devices is currently a growing trend among users, that is causing a demand for higher bandwidth and better provisioning throughout the network infrastr...
Conference Paper
Full-text available
Video streaming on mobile devices is prone to a multitude of faults and although well established video Quality of Experience (QoE) metrics such as stall frequency are a good indicator of the problems perceived by the user, they do not provide any insights about the nature of the problem nor where it has occurred. Quantifying the correlation betwee...
Conference Paper
Full-text available
A significant fraction of Internet traffic is now encrypted and HTTPS will likely be the default in HTTP/2. How- ever, Transport Layer Security (TLS), the standard protocol for encryption in the Internet, assumes that all functionality resides at the endpoints, making it impossible to use in-network services that optimize network resource usage, im...
Article
Full-text available
A significant fraction of Internet traffic is now encrypted and HTTPS will likely be the default in HTTP/2. However, Transport Layer Security (TLS), the standard protocol for encryption in the Internet, assumes that all functionality resides at the endpoints, making it impossible to use in-network services that optimize network resource usage, impr...
Article
Through their normal operation, cellular networks are a repository of continuous location information from their subscribed devices. Such information, however, comes at a coarse granularity both in terms of space, as well as time. For otherwise inactive devices, location information can be obtained at the granularity of the associated cellular sect...
Conference Paper
Increased user concern over security and privacy on the Internet has led to widespread adoption of HTTPS, the secure version of HTTP. HTTPS authenticates the communicating end points and provides confidentiality for the ensuing communication. However, as with any security solution, it does not come for free. HTTPS may introduce overhead in terms of...
Technical Report
Full-text available
This deliverable presents an extended set of Analysis Modules, including both the improvements done to those presented in deliverable D4.1 as well as the new analysis algorithms designed and developed to address use-cases. The deliverable also describes a complete workflow description for the different use-cases, including both stream processing fo...
Article
The Internet's universality is based on its decentralization and diversity. However, its distributed nature leads to operational brittleness and difficulty in identifying the root causes of performance and availability issues, especially when the involved systems span multiple administrative domains. The first step to address this fragmentation is...
Technical Report
Full-text available
This public deliverable describes the design and specifica of a first set of basic analysis modules for addressing the use cases iden in WP1. The document focuses on the required algorithms, which use as input the measurements and analysis pro-vided by the lower layers (WP2 and WP3) of the mPlane architecture to provide more advanced analysis and a...

Citations

... Dyno [22] raised the priority of key frames to provide differentiation for different data in the learning process. D3 [23] proposed a dynamic DNN decomposition system for synergistic inference without precision loss. ...
... NPUs) to efficiently run DL workloads [3]. These often come in various configurations in terms of their compute/memory capabilities and power envelopes [4] and co-exist in the wild as a rich multi-generational ecosystem (system heterogeneity) [74]. These devices bring intelligence through users' interactions, also innately heterogeneous amongst them, leading to non-independent or identically distributed (non-IID) data in the wild (data heterogeneity). ...
... Spurred by that, very recently, there have been a few approaches targeting the use of mixed-precision for representing DNN datastructures to design more efficient quantized DNNs [49]- [52]. This flexibility was originally not supported by chip vendors until recently when some advanced chips were released, including Apple A12 Bionic chip, Nvidia's Turing GPU architecture, and Imagination neural network accelerator IP, all of these support mixed-precision arithmetic [53]. Besides industry, academia also is working on bit-level flexible hardware design to accelerating DNNs. ...
... It uses multiple early-exits with the usual pooling and dense layer connections. Leontiadis et al. [35] builds hierarchical models from a global base model for mobile edge computing. The appropriate model is selected at runtime depending on the computation budget to exit early. ...
... Early-exit networks have successfully been deployed in a multitude of tasks, spanning from image classification [12], [13] to semantic segmentation [14] or even various NLP tasks [15]. On the system side, existing work has so far focused on the hardware-aware design of the exit policy, and the placement and architecture of the exits along the depth of the backbone network [12]- [14], the distributed execution of such models across device and server [16], [17] and the codesign of an early-exit model and its hardware accelerator [18]. ...
... In recent years, deep learning (DL) models have emerged as the state-of-the-art in several AI inference tasks. Ranging from the recognition of objects [1] and emotions [2] to scene [3] and speech understanding [4], their unrivalled accuracy has made deep neural networks (DNNs) an enabler of many mobile apps. Currently, developers who seek state-of-the-art accuracy and wide device compatibility typically resort to offloading the DL model execution to a remote server [5]. ...
... A client-like host usually acts as the downstream user of a server, and it performs ML based on the ML framework/algorithm and models provided by the server. Common use cases are i) on-device inference (e.g., [116]), the client conducts predictions on its own data with another individual's model, ii) on-device personalization (e.g., [117]), the client personalizes the model based on its data for later inference, and iii) federated learning (e.g., [22]), the client trains the model to further contribute a global model owned by another party (e.g., server). In all these cases, the client owns the data; thus, this client's TEE does not need to hide these local data from itself. ...
... These online collaborative activities can happen for interactions between online communities with particular purposes (Zannettou et al. 2017;Kumar et al. 2018;Tan 2018;Mariconti et al. 2019). Zannettou et al. (2017) analyzed the interaction among communities in different online platforms and showed that news shared in Twitter can affect its spread in 4chan and Reddit. ...
... Prior research into toxic comments has focused on a variety of themes including the experiences of targets [39,40,44], the characterizations of specific, large-scale events like #GamerGate [8]; early warnings for how toxic conversations escalate [47,49,50], the off-platform coordination tactics for raiding and calls to incite attacks against targets [1,30], and the impact of intervention techniques such as banning accounts or entire communities [5,7,36]. While these studies all paint a rich tapestry of toxic behaviors that occur online, none capture the long-term activities of abusive accounts (i.e., accounts that post toxic comments), such as their toxicity behaviors or their impact on the platform itself. ...
... This highlights that the nature of the problem lies between the modelling social dynamics and proxemics values which capture distance and orientation information to some extent, represented by [29]. In the case of [25], data from motion sensors (accelerometer and gyroscope) were also incorporated with proximity features for multimodal detection of groups using smart phones. Other custom sensors have been developed to measure proximity, relative orientation, motion, and/or a combination thereof (e.g., light tags [30], Rhythm badges [28] and the Midge [1]). ...