Previous research has shown that abusers in an intimate relationship can find plenty of technical advice, tools, and how-to guides online for covertly conducting intimate partner surveillance (IPS). However, it is unclear what resources survivors seeking to defend themselves against IPS can use. To address this gap, we first conducted a survey-based study with 63 survivors recruited via Prolific to understand what resources survivors rely on. We showed that 45% utilized online resources for assistance, with 67% of them relying on search engines. We then conducted a systematic survey of the results obtained via Google search engine to identify resources available for survivors. We found that the resources survivors can find online contain poor, inaccurate, and unactionable advice. They are hard to understand and do not help mitigate IPS. To investigate whether the lack of useful resources is solely experienced by survivors, we also crawled resources that abusers will find online. We found that abusers can easily find resources recommending spyware apps and hidden devices and often explicitly promoting IPS. We also compared the understandability and actionability of the resources using an adopted Patient Education Materials Assessment Tool (PEMAT) score. We concluded that resources available to abusers are significantly more understandable and actionable than those available to survivors.

Secure multi-party computation (MPC) techniques can be used to provide data privacy when users query deep neural network (DNN) models hosted on a public cloud. State-of-the-art MPC techniques can be directly leveraged for DNN models that use simple activation functions (AFs) such as ReLU. However, these techniques are ineffective and/or inefficient for the complex and highly non-linear AFs used in cutting-edge DNN models. We present Compact, which produces piece-wise polynomial approximations of complex AFs to enable their efficient use with state-of-the-art MPC techniques. Compact neither requires nor imposes any restriction on model training and results in near-identical model accuracy. To achieve this, we design Compact with input density awareness, and use an application specific simulated annealing type optimization to generate computationally more efficient approximations of complex AFs. We extensively evaluate Compact on four different machine-learning tasks with DNN architectures that use popular complex AFs silu, gelu, and mish. Our experimental results show that Compact incurs negligible accuracy loss while being 2x-5x computationally more efficient than state-of-the-art approaches for DNN models with large number of hidden layers. Our work accelerates easy adoption of MPC techniques to provide user data privacy even when the queried DNN models consist of a number of hidden layers, and trained over complex AFs.