Get Fact Up
12 subscribers
13 photos
9 videos
64 links
Join TNP Hosts Denver Attaway and Yona Aniwodi for some good researching & conversation every Thursday at 9PM ET.
Download Telegram
The potatoes and butter nut squash is still roasting with the chickens. Dr Lawrence was a parish leader, grandfather, great grand father and father and we lived at there house during the blizzard of 93 when we were without power for 16 days.

He was the best of men and will be missed. 😔
❤3
wowzerz
Benjamin IP Rubinstein, Blaine Nelson, Ling Huang, Anthony D Joseph, Shing-hon Lau, Satish Rao, Nina Taft, and J Doug Tygar. 2009.
Antidote: understanding and defending against poisoning of anomaly detectors. In Proceedings of the 9th ACM SIGCOMM Conference on Internet Measurement.
https://www.semanticscholar.org/paper/ANTIDOTE%3A-understanding-and-defending-against-of-Rubinstein-Nelson/1bd10813ade534b5500e92600d909bacb514138d
Huang Xiao, Battista Biggio, Gavin Brown, Giorgio Fumera, Claudia Eckert, and Fabio Roli. 2015.
Is feature selection secure against training data poisoning?. In International Conference on Machine Learning.
Chaofei Yang, Qing Wu, Hai Li, and Yiran Chen. 2017b.
Generative Poisoning Attack Method Against Neural Networks.
arXiv preprint 1703.01340 (2017). https://arxiv.org/pdf/1703.01340
E. The Loss-based Countermeasure against Poisoning Attack
Moreover, we propose an universal method to detect the
aforementioned poisoning attack methods, which is shown in
Algorithm 3. Once a data (no matter normal or poisoned) is
injected into the target model, the loss of the target model is
recorded. If the loss exceeds the pre-determined threshold Lth,
a warning will show up. If the number of warnings exceeds the
threshold Wth, the accuracy check will be triggered to examine
if a poisoning attack is indeed being conducted. Our method
is based on the phenomenon that the poisoned inp
nput and label
pair usually results in a larger loss, compared with the normal
one. This is naturally understandable as the goal of poisoning
attack is to minimize the model accuracy, which is realized
through maximizing the loss. The poisoned data alters target
model’s original decision boundary. On the contrary, normal
data stays inside the decision region (at least most of time),
inducing a relatively smaller loss. We are able to monitor the
condition of inputs by checking the loss periodically.
H. Xiao, H. Xiao, and C. Eckert. Adversarial label flips attack on support
vector machines. European Conference on Artificial Intelligence, pages
870–875, 2012.
friends of the coal company = enemies of the coal miner. basic equation
woohooo preshow starts now
🎄1
Update on that bit from Niger - how the Pres/P.M. said they were about to be attacked b y USA, Izreal, France and Ukraine connected forces?