An approach to combating network intrusion is the development of systems applying machine learning and data mining techniques. Many IDS (Intrusion Detection Systems) suffer from a high rate of false alarms and missed intrusions. We want to be able to improve the intrusion detection rate at a reduced false positive rate. The focus of this paper is rule- learning, using RIPPER, on highly imbalanced intrusion datasets with an objective to improve the true positive rate (intrusions) without significantly increasing the false positives. We chose RIPPER to induce the comprehensibility in the model as is required by humans using the system. To counter imbalance in data, we implement a combination of oversampling (both by replication and synthetic generation) and undersampling techniques. We also propose a clustering based methodology for oversampling by generating synthetic instances. We evaluate our approaches on two intrusion datasets — destination and actual packets based — constructed from actual Notre Dame traffic, giving a flavor of real-world data with its idiosyncracies. Using ROC analysis, we show that oversampling by synthetic genera- tion of minority (intrusion) class outperforms oversampling by replication and RIPPER's loss ratio to inherently counter class imbalance. Additionaly, we establish that our clustering based approach is more suitable for the detecting intrusions.

The availability of “always-on” communications has tremendous implications for how people interact socially. In particular, sociologists are interested in the question if such pervasive access increases or decreases face-to-face interactions. Unlike triangulation which seeks to precisely define position, the question of face-to-face interaction reduces to one of proximity, i.e., are the individuals within a certain distance? Moreover, the problem of proximity estimation is complicated by the fact that the measurement must be quite precise (1-1.5 m) and can cover a wide variety of environments. Existing approaches such as GPS and Wi-Fi triangulation are insufficient to meet the requirements of accuracy and flexibility. In contrast, Bluetooth, which is commonly available on most smartphones, provides a compelling alternative for proximity estimation. In this paper, we demonstrate through experimental studies the efficacy of Bluetooth for this exact purpose. We propose a proximity estimation model to determine the distance based on the RSSI values of Bluetooth and light sensor data in different environments. We present several real world scenarios and explore Bluetooth proximity estimation on Android with respect to accuracy and power consumption.

Health wearables in combination with gamification enable interventions that have the potential to increase physical activity—a key determinant of health. However, the extant literature does not provide conclusive evidence on the benefits of gamification, and there are persistent concerns that competition-based gamification approaches will only benefit those who are highly active at the expense of those who are sedentary. We investigate the effect of Fitbit leaderboards on the number of steps taken by the user. Using a unique data set of Fitbit wearable users, some of whom participate in a leaderboard, we find that leaderboards lead to a 370 (3.5%) step increase in the users’ daily physical activity. However, we find that the benefits of leaderboards are highly heterogeneous. Surprisingly, we find that those who were highly active prior to adoption are hurt by leaderboards and walk 630 fewer steps daily after adoption (a 5% relative decrease). In contrast, those who were sedentary prior to adoption benefited substantially from leaderboards and walked an additional 1,300 steps daily after adoption (a 15% relative increase). We find that these effects emerge because sedentary individuals benefit even when leaderboards are small and when they do not rank first on them. In contrast, highly active individuals are harmed by smaller leaderboards and only see benefit when they rank highly on large leaderboards. We posit that this unexpected divergence in effects could be due to the underappreciated potential of noncompetition dynamics (e.g., changes in expectations for exercise) to benefit sedentary users, but harm more active ones. This paper was accepted by Chris Forman, information systems. Funding: This work was supported by the National Institutes of Health [Grant 5R01HL117757]. Supplemental Material: The data and online appendix are available at https://doi.org/10.1287/mnsc.2022.4581 .

Background: Stress can have adverse effects on health and well-being. Informed by laboratory findings that heart rate variability (HRV) decreases in response to an induced stress response, recent efforts to monitor perceived stress in the wild have focused on HRV measured using wearable devices. However, it is not clear that the well-established association between perceived stress and HRV replicates in naturalistic settings without explicit stress inductions and research-grade sensors. Objective: This study aims to quantify the strength of the associations between HRV and perceived daily stress using wearable devices in real-world settings. Methods: In the main study, 657 participants wore a fitness tracker and completed 14,695 ecological momentary assessments (EMAs) assessing perceived stress, anxiety, positive affect, and negative affect across 8 weeks. In the follow-up study, approximately a year later, 49.8% (327/657) of the same participants wore the same fitness tracker and completed 1373 EMAs assessing perceived stress at the most stressful time of the day over a 1-week period. We used mixed-effects generalized linear models to predict EMA responses from HRV features calculated over varying time windows from 5 minutes to 24 hours. Results: Across all time windows, the models explained an average of 1% (SD 0.5%; marginal R2) of the variance. Models using HRV features computed from an 8 AM to 6 PM time window (namely work hours) outperformed other time windows using HRV features calculated closer to the survey response time but still explained a small amount (2.2%) of the variance. HRV features that were associated with perceived stress were the low frequency to high frequency ratio, very low frequency power, triangular index, and SD of the averages of normal-to-normal intervals. In addition, we found that although HRV was also predictive of other related measures, namely, anxiety, negative affect, and positive affect, it was a significant predictor of stress after controlling for these other constructs. In the follow-up study, calculating HRV when participants reported their most stressful time of the day was less predictive and provided a worse fit (R2=0.022) than the work hours time window (R2=0.032). Conclusions: A significant but small relationship between perceived stress and HRV was found. Thus, although HRV is associated with perceived stress in laboratory settings, the strength of that association diminishes in real-life settings. HRV might be more reflective of perceived stress in the presence of specific and isolated stressors and research-grade sensing. Relying on wearable-derived HRV alone might not be sufficient to detect stress in naturalistic settings and should not be considered a proxy for perceived stress but rather a component of a complex phenomenon.

Personalized predictions have shown promises in various disciplines but they are fundamentally constrained in their ability to generalize across individuals. These models are often trained on limited datasets which do not represent the fluidity of human functioning. In contrast, generalized models capture normative behaviors between individuals but lack precision in predicting individual outcomes. This paper aims to balance the tradeoff between one-for-each and one-for-all models by clustering individuals on mutable behaviors and conducting cluster-specific predictions of psychological constructs in a multimodal sensing dataset of 754 individuals. Specifically, we situate our modeling on social media that has exhibited capability in inferring psychosocial attributes. We hypothesize that complementing social media data with offline sensor data can help to personalize and improve predictions. We cluster individuals on physical behaviors captured via Bluetooth, wearables, and smartphone sensors. We build contextualized models predicting psychological constructs trained on each cluster's social media data and compare their performance against generalized models trained on all individuals' data. The comparison reveals no difference in predicting affect and a decline in predicting cognitive ability, but an improvement in predicting personality, anxiety, and sleep quality. We construe that our approach improves predicting psychological constructs sharing theoretical associations with physical behavior. We also find how social media language associates with offline behavioral contextualization. Our work bears implications in understanding the nuanced strengths and weaknesses of personalized predictions, and how the effectiveness may vary by multiple factors. This work reveals the importance of taking a critical stance on evaluating the effectiveness before investing efforts in personalization.