Last month, the San Francisco Board of Supervisors voted in favor of allowing that city's police department to deploy robots equipped with a potential to kill, should a situation--in the estimation of police officers--call for lethal force. With that decision, the board appeared to have delivered the city to a dystopian future. The vote garnered a loudly negative response from the public, and this week the supervisors reversed course and sent the policy back to committee. But the fact that the decision initially passed--and may yet pass in some form--should not have been surprising. Police departments around the country have been acquiring robotic devices for decades.

Last week, we talked about killer robots. That piece was inspired by a proposal that would allow San Francisco police to use robots for killing "when risk of loss of life to members of the public or officers is imminent and outweighs any other force option available to SFPD." Last night, that proposal passed the city's board of supervisors with an 8-3 vote. The language was included in a new "Law Enforcement Equipment Policy" filed by the San Francisco Police Department in response to California Assembly Bill 481, which requires a written inventory of the military equipment utilized by law enforcement. The document submitted to the board of supervisors includes -- among other things -- the Lenco BearCat armored vehicle, flash-bang grenades and 15 submachine guns.

Schuyler Moore, Chief Technology Officer at US Central Command (CENTCOM), said Wednesday that such ideas are expected to play a much bigger role in bolstering American troops. CENTCOM is the US combatant command that covers the Middle East and parts of northern Africa and southern Asia. "This tool is intended to do is to provide the opportunity for the boots on the ground that are closest to the threat and closest to their operational environment to mimic that environment as accurately as possible and to, as well, deploy threats against their base and to train against threats on their base that they see and that they interact with sometimes daily," said army Sergent Mickey Reeve, who developed the counter-unmanned aerial system training software. He added that the US will use the Red Sands Integrated Experimentation Center in Saudi Arabia to "pressure test" the new technology, but it will be moving around the region. Sergeant Reeve noted the technology is not tailored solely towards Iranian-made drones, "but it was built to emulate any sort of Unmanned Aerial Systems."

STR's Analytics Division researches and develops novel technologies to solve challenging national security problems through advanced analytics. Our team consists of passionate and motivated engineers and scientists with advanced degrees in engineering, computer science, mathematics, physics, and data science. We use our expertise and creativity to take innovative ideas from conception to mature implementation in order to improve mission success of our customers. The Signals Exploitation and Tracking (SET) Group in the Analytics Division focuses on applying machine learning, statistics, estimation, and information theory algorithms for signals exploitation, target tracking, and system resource management. As a Research Scientist at STR, you will help develop disruptive technologies focused on signals exploitation, estimation theory, system resource management, and systems analysis.

Neuralink is reportedly being investigated by the US government for possibly mistreating animals in lab experiments as the company rushes to build an implantable brain chip. The startup, founded in 2016 by belligerent biz baron Elon Musk, is developing a medical device to help people afflicted with brain disorders to communicate, see, or move more easily. The Tesla tycoon said he wants to see the chips in humans next year. However, before Neuralink obtains permission to start conducting trials on humans from the Food and Drug Administration (FDA), it has to test the technology on animals to show it's safe enough for people to try. A lawsuit, filed in February by animal rights group the Physicians Committee for Responsible Medicine, accused the startup of killing monkeys after a series of sloppy surgical procedures to insert electrodes inside their skulls.

More than three years ago, this editor sat down with Sam Altman for a small event in San Francisco soon after he'd left his role as the president of Y Combinator to become CEO of the AI company he co-founded in 2015 with Elon Musk and others, OpenAI. At the time, Altman described OpenAI's potential in language that sounded outlandish to some. Altman said, for example, that the opportunity with artificial general intelligence -- machine intelligence that can solve problems as well as a human -- is so incomprehensibly enormous that if OpenAI managed to crack it, the outfit could "maybe capture the light cone of all future value in the universe." He said that the company was "going to have to not release research" because it was so powerful. Asked if OpenAI was guilty of fear-mongering -- Elon Musk, a co-founder of the outfit, has repeatedly called all organizations developing AI to be regulated -- Altman talked about dangers of not thinking about "societal consequences" when "you're building something on an exponential curve."

Anyone who's seen the show knows that this is not what happened. Instead, it's a somewhat humorous misconfiguration of the details--sorta like listening to a friend misremember a series that they haven't seen in awhile. In a different conversation, I asked ChatGPT how the TV show Gilligan's Island ended. In reality, Gilligan's Island was cancelled by its network, so there was no ending.

The developments of satellite communication in network systems require strong and effective security plans. Attacks such as denial of service (DoS) can be detected through the use of machine learning techniques, especially under normal operational conditions. This work aims to provide an interruption detection strategy for Low Earth Orbit (\textsf{LEO}) satellite networks using deep learning algorithms. Both the training, and the testing of the proposed models are carried out with our own communication datasets, created by utilizing a satellite traffic (benign and malicious) that was generated using satellite networks simulation platforms, Omnet++ and Inet. We test different deep learning algorithms including Multi Layer Perceptron (MLP), Convolutional Neural Network (CNN), Recurrent Neural Network (RNN), Gated Recurrent Units (GRU), and Long Short-term Memory (LSTM). Followed by a full analysis and investigation of detection rate in both binary classification, and multi-classes classification that includes different interruption categories such as Distributed DoS (DDoS), Network Jamming, and meteorological disturbances. Simulation results for both classification types surpassed 99.33% in terms of detection rate in scenarios of full network surveillance. However, in more realistic scenarios, the best-recorded performance was 96.12% for the detection of binary traffic and 94.35% for the detection of multi-class traffic with a false positive rate of 3.72%, using a hybrid model that combines MLP and GRU. This Deep Learning approach efficiency calls for the necessity of using machine learning methods to improve security and to give more awareness to search for solutions that facilitate data collection in LEO satellite networks.

Deep Reinforcement Learning (RL) agents are susceptible to adversarial noise in their observations that can mislead their policies and decrease their performance. However, an adversary may be interested not only in decreasing the reward, but also in modifying specific temporal logic properties of the policy. This paper presents a metric that measures the exact impact of adversarial attacks against such properties. We use this metric to craft optimal adversarial attacks. Furthermore, we introduce a model checking method that allows us to verify the robustness of RL policies against adversarial attacks. Our empirical analysis confirms (1) the quality of our metric to craft adversarial attacks against temporal logic properties, and (2) that we are able to concisely assess a system's robustness against attacks.

Although machine learning based algorithms have been extensively used for detecting phishing websites, there has been relatively little work on how adversaries may attack such "phishing detectors" (PDs for short). In this paper, we propose a set of Gray-Box attacks on PDs that an adversary may use which vary depending on the knowledge that he has about the PD. We show that these attacks severely degrade the effectiveness of several existing PDs. We then propose the concept of operation chains that iteratively map an original set of features to a new set of features and develop the "Protective Operation Chain" (POC for short) algorithm. POC leverages the combination of random feature selection and feature mappings in order to increase the attacker's uncertainty about the target PD. Using 3 existing publicly available datasets plus a fourth that we have created and will release upon the publication of this paper, we show that POC is more robust to these attacks than past competing work, while preserving predictive performance when no adversarial attacks are present. Moreover, POC is robust to attacks on 13 different classifiers, not just one. These results are shown to be statistically significant at the p < 0.001 level.