Tuesday, January 15, 2019

Blog Post 1 - The Future and Security of AI and Computing

As time goes on, we as a society see a push for more connected technologies and services and the integration of AI into every facet it can be integrated. There are obvious benefits to this as there you will be able to access all of your data and files from anywhere, at anytime and automate a lot of your daily tasks. AI are seeing an increase in popularity in cell phone and tech assistants and even in self-driving cars. While all the connectivity is beneficial, it largely increases both security risks and privacy issues. I feel that the future will bring an increase in security risks that could outweigh the benefits of connectivity.

AI is definitely in the spotlight currently as self driving cars gain traction and the design of general use AI are becoming more popular. The benefits of AI could be numerous but the risks the impose are also numerous. In an article from September 2018, the ease at which AI could be fooled or hacked was discussed. The article talked about the astonishingly easy methods AI could be convinced to see or hear something different than what it was actually exposed to. A frightening example is that some AI researchers "put some stickers on a stop sign, fooling a common type of image recognition AI into thinking it was a 45-mile-per hour speed limit sign." [1] This is very bad as it would be so easy for someone to fool a self-driving car like that and to cause a serious accident. The article discusses that this type of redirect can be done not only visually but also for auditory AI. The same group of researchers also came up with a "trick that makes an image recognition AI not only mislabel things, but hallucinate them." [1] AI developers are aware of some of these tricks and have begun to come up with countermeasures but other AI researchers needed only a couple of days to circumvent those measures again.

Hacking is prevalent in not just AI but in all aspects of computing and the internet as evidenced by the news storied every so often about a company being hacked (Sony being a big one). In an interview with Greg Shannon, a former chief scientist at Carnegie-Mellon's Software Engineering Institute and is currently the assistant director for cybersecurity strategy at the White House Office of Science and Technology Policy, he stated that the reason that hacking and security breaches are so lucrative right now is that "what's available online, and its value, keep increasing exponentially- and so do the incentives to exploit systems and steal data." [2] These incentives will only grow as connectivity becomes more encompassing. This is concerning to me given the rise of AI. As mentioned above, it can be very simple to hack an AI or change what it perceives. This vulnerability could be used as the point of easiest access for hackers to gain access to private and important data. Shannon's suggestion on making everything more secure lies in the Internet of Things (IoT). He believes that since the IoT is still in it's infancy, we can build security into the system from the beginning so that in the future it is a core part of technology.

Another potential solution I feel lies in edge computing. The premise of edge computing is "computing that's done at or near the source of the data, instead of relying on the cloud atone of a dozen data centers to do all the work." [3] This is a potential solution as as described in the article, having edge computing in your devices in the future, a lot of things will be managed centrally and pushed to these devices. This will help security as you are not at risk of a security breach from the cloud as each device will be updated and secured on its own or through updates from the developer. A potential downside is discussed in the article that since the developers will push updates out and the user plays no part in the installation, consumers would be at the whim of whatever the developers wanted to put onto the device, whether we like it or not. We may gain better security for our important data, but may lose privacy and agency in our devices.

References
[1] Hutson, Matthew. "Hackers Easily Fool Artificial Intelligences." Science, Vol 361, Issue 6399, pp               215, 20 Jul 2018 <http://science.sciencemag.org/content/361/6399/215.full> 14 Jan 2019.
[2] Talbot, David. "Why We're So Vulnerable" MIT Technology Review. 25 Jan 2016
          <https://www.technologyreview.com/s/545621/why-were-so-vulnerable/> 14 Jan 2019.
[3] Miller, Paul. "What is edge computing?" The Verge, 7 May 2018
          <https://www.theverge.com/circuitbreaker/2018/5/7/17327584/edge-computing-cloud-google-              microsoft-apple-amazon> 14 Jan 2019.

Comments on Other Posts

To Blas Andres Rodriguez Vieria
I agree that there are some aspects of building life and management that can not be controlled or modeled in BIM. Weather and communication are critical in building management and cannot be overlooked. If new technology were created so that that information could be monitored in BIM then building management would reach a whole new era of usefulness. BIM could be used even more in-depth to control a building. It could be used to create more efficient, intelligent buildings as certain measures could be controlled based on the BIM model and data.

To Julie Hansted Andersen
I agree that the phenomenon of people fearing that AI will come for their jobs is an interesting problem. I can see why people would be concerned about that as many jobs would not be as necessary with the use of AI. The Holberton school is an interesting concept that I think could be very beneficial. Since tuition is often the reason people are unable to attend higher education, delaying the payment until a job is secured and basing the tuition off of that salary could be a great way to get around that problem. I think that this type of school and payment would be very beneficial even today without the presence of AI taking jobs.

To Adam Raifsnider
I cannot agree more that the HVAC system and energy use of a building need to be considered from the genesis of design. It is the most costly to maintain over time so the more efficient and effective it is, the better. I agree that HVAC sensors and intelligence are crucial for the future to ensure that buildings operate to the best of their ability. The Keen Smart Vent is very interesting and I believe that more technologies such as that need to be introduced to the smart environment.

2 comments:

  1. Richard,
    I agree that while AI provides great convenience and its connectivity has a clear benefits, the security threat AI presents needs to be thoroughly investigated and protected against exploitation. AI presents a very dangerous relationship between convenience and safety, and it is clear, especially with the notes on hackers fooling the AI used in self driving cars to drive at higher speeds, that we must be careful that systems do not become so overly automated that safety is compromised. The idea that hacking and security breaches are becoming more popular due to the lucrative information that is stored online is alarming when considering all of the future tech that has interconnected communication systems. Edge computing is an interesting solution, but again, it becomes an issue of loss of privacy for security. The privacy is up to the developer to be benevolent and I imagine a lot of people will have a hard time trusting developers, especially with all of the Facebook privacy developments recently.

    ReplyDelete
  2. The industry of cyber security is growing, and has been for years. I like the articles you chose and the points you made, as it draws attention to this, and the fact that even with this growing industry and advancing the technology that goes into cyber security, it is still vulnerable to attacks and data breaches. Similarly to your point with the self-driving cars, knowing how a system works from the inside can often times make it easy to infiltrate and steal from, whether that system manifests itself physically in a car hooked up to numerous sensors, or a digital system protected by code. Both have vulnerabilities, but both can be protected, given the proper set-up work, as per Shannon’s suggestion in your third paragraph.

    ReplyDelete

Note: Only a member of this blog may post a comment.