AI for evil. #EuroCyberWeek

When we use AI for evil. threat categories. #EuroCyberWeek pic.twitter.com/0rmm66GZ7b

— 𝙻𝚞𝚒𝚜 𝙺𝚛 (@lcsuarezt) 19 novembre 2018

C&ESAR 2018, day 1

During this week, Rennes will be hosting the European Cyber Week. This is a great opportunity for communication service providers (CSP), manufacturers, industry players and end users to meet and present current research activities about Artificial Intelligence and cyber security.

Opening speeches stressed on the importance on AI for defense, since each day networks and services experience attacks, not only from external actors, but from inside our networks. The idea is not only to disrupt service, but to steal data, jeopardizing research and intellectual property. This constitutes the rise of adversarial AI, and to react to this menace, it is necessary to enhance our understanding and usage of generative adversarial networks.

Other interesting topic is about "opening the black box" in order to have an explanatory AI, which has the goal of explaining the reasoning of the decisions taken by AI. This is important to gain confidence of users and to speed up the adoption of these techniques in various use cases.

Some other topics that caught my attention:

• The diverse malicious AI techniques to break cyber-security, such as data poisoning. This in order to induce errors on the machine learning model.
• The bias problem, which most of the times has its source of the training data and leads to erroneous decisions at the end of the AI process.
• The usage of different frameworks to have behavioral analysis, very useful to detect deviation of usual usage patterns of an entity. This helps to detect compromised entities our users that deviate or abuse of privileges.

I hope day 2 brings new insights and interesting approaches to better understand AI and its challenges.

European Cyber Week à Rennes: Cybersecurity of Internet of Things

The last session of the European Cyber Week had a focus on IoT .

The key messages in this session were:

• Even though that the approach to security has to be end-to-end, each layer of the SOA for IoT (sensing layer, network layer, service layer, interface layer) must try to enforce its own security mechanism. Each layer as its own weaknesses.
• The security mindset has to be used since the conception of the service. As an example, the election of the sensors and actuators plays an important role, because they are manufactured by third party companies that seek economy, fast delivery, sell millions of devices at a very low cost. It is common that their security mechanisms on those systems on chips are not the best. 
• Since R8, 3GPP has pushed evolution towards LTE-A and LTE-A Pro. The standardization entity has designed mechanisms to enable IoT systems to reduce power consumption, expose services via Service Capability Exposure Function (SCEF)... and well, 5G seeks to enhance escalability by providing the same architecture no matter the radio access technology of the IoT system. The new generation core would receive traffic from heterogeneous access technologies.
• For a service operator, the cloud plays a key role, not only as a "place" where data is received, processed and stored, but as a central intelligence analysis center in order to detect anomalous activity and deploy countermeasures: detecting evil behavior is necessary, but trying to dynamically deploy the defense mechanism is also important. 
• Since at the end all is measured with numbers (costs and return of investment), there is a trade-off between how exigent is the SLA provided (availability, integrity and confidentiality) for the data and the analysis of the risk of being compromised. I think that the parameters of risk could be the value of the data, how important it is for the business, the additional latency and decreased battery life for a sensor when a full security scheme is deployed. Parameters are countless. 
• I think that some of the challenges for IoT regarding security (scalability, interoperability, management, security and privacy) can be addressed via the network slicing concept. This by providing means to escalate the network resources as needed and by deploying IDS/IPS functions on-demand, where needed. The isolation that a network slice provides could enable the contention of an attack by creating network honeypots or quarantine slices to contain malicious or suspicious activity. Feedback loops and OSS/BSS interaction is important to achieve this.

From my point of view, the service operators have no awareness of where is the service implemented inside the cloud. they are blind to the physical location of the functions. And well, should they worry? they just need the service, don't care how it is implemented. For them, something important is that the SLA promised by the cloud provider are accomplished properly. It is inside a network slice? it is in a physical dedicated server? it is up to the communications service provider to make the decision and deploy the best infrastructure for the customer.

This whole week was amazing. Has given me great view of the importance of security for industry and all the requirements that an architecture should support in order to meet the exigencies of the (new) use cases that industry and other actors would be implementing.

European Cyber Week à Rennes: Cybersecurity and healthcare focus day

For me, the key conference today was The Future of Healthcare – Scotland: paving the way, by Pr. William (Bill) BUCHANAN (@billatnapier). This conference blew my mind away.

It is known for all of us that the current technologies that we use to access Internet and the services built over it, use inventions developed more than 30 years ago. Improvements to the web, IP and TCP have been made by patching new features over them or creating new layers over them to provide new functions. And it is not only IP, or TCP: this involves also the use of STP on L2 networks, or BGP as a protocol used on Internet route announcement (and now inside data centers). Attention please, those layers are not like abstraction layers, but layers that obscure and makes difficult the operation of the system. We are using the same old tools to try to fix new problems envisioned by the massive growth of the Internet ecosystem and all the services that are supported. As suggested by Prof. Buchanan, a master reset should be done, in order to re-create the foundations of Internet having a security framework in mind.  But so far, we know this is very difficult. Just check out the example of IPv6, in which its adoption has had a low pace, because of the lack of incentive$ for telecoms to fully deploy it. All is about businesses and the return of investment after a change in a network. 

The key message here is that we are in the middle of a great opportunity to “make things right” in the foundation of the standardization process for 5G. Learn from the old technologies, have a security and privacy mindset in the implementation. The network may be seen by others like a dumb pipe, but the communication service operators are the aggregators of those pipes, and have to make sure that those pipes behave well and do not mess one with another. On the other hand, the deployment of (virtual) network functions and applications should be taken care too: this because 5G relies on software to provide functionality and is easily corruptible. How to make sure that the (virtual) network function is issued by a trusted entity? How I make sure that the orders from OSS/BSS are legitimate?

Assorted ideas in my conversation with Prof. Buchanan: encryption should be made on the source of data generation, or near the edge. The user must have the power over the data. The network must provide the necessary QoS according to the type of data, but  has to be agnostic to the content: the operator must not know about the data.

On the other side, since this focus day is centered on healthcare, convinces me once more on my quest to have a human approach of technology: to empower, enable him to pursue its dreams, to protect  the user and his data. This way, people will trust the system.

European Cyber Week à Rennes, day three

My key points for the last day of the C&ESAR conferences:

• The conference I enjoyed the most was "une autre vision de lq Cyber Threat Intelligence (CTI) by Airbus Cybersecurity. The motive is simply the reasoning behind the threat model and the cyberdefense strategy they showed and the during the presentation. 
• Regarding the threat model, covers all the steps from motivation to perform harm, compromising the target and exploiting the vulnerability.
• About the cyberdefense strategy, covered a complete approach with:
• Strategy: administrative decisions on the defense.
• Conception: use the imagination! Architects propose a solution to the security problems.
• Tactic: the defense. How are we going to defend? Made by the security engineers. How to correlate?
• Operation: BAU. SoC, CSIRT. Technical formation to personnel.  
• This gave me a lot to think about, because having a clear threat model allows to to have a vision of how to check an architecture for weak links and possible solutions to those potential problems. At some point, we would like to have granularity of the (virtual) network functions in order to have a flexible service composition and simple lightweight functions firing up when necessary. But the problem is the multiplication of the points of failure that are created. 
• An well, securing all those points of failures have costs in terms of money, processing time, memory, delay and latency... It is a trade off with the value of what I want to protect. What is the justification for such an investment?

The afternoon session was developed in Secure-IC. The topic was about the business of digital security. The subject was a little bit more administrative to my taste (or to my interest). Some isolated comments:

• Europe has no representative in the top 10 industries in the world: first 8 are USA, last 2 of the top 10 are Chinese.
• It is a shame that all is shaped by politics, being technology also affected by this.
• 90% of advertisement in the world is captured by Google and Facebook.
• Among the technical priorities in DGA plan, they want:
• Evaluation and orientation of COTS technologies.
• Improve architecture and the resilience of large systems (ships, aircraft…) taking into account the operational constraints. 
• LOL, this sounds like they are sharing some of the functional needs of 5G along with its enabling technologies. It is a fact that SDN and NFV would help to achieve these requirements. I am imagining right now network slices for ships, aircraft, hospitals, smart cities. In fact, as the speaker said, a ship, for example, is like a smart city! has its own energy source, water supply control, temperature control, CCTV, the crew.. a small scale city.

So far, I have more ideas, more questions, more reading to do and so much to learn; got to keep going.

Faire la pause: European Cyber Week à Rennes, day two

On day two, the approach was quite different but no less interesting: the topics covered training, penetration testing and protection from treats. Key points:

• Simulation environments are very important because of the several use cases, for example, you could use a simulation to recreate an attack by leveraging on virtualization and traffic generators to replay the packets and perform a post-mortem analysis. Other use is for training by using a virtualised version of the real products, topology and traffic generators and controls to provide a learning environment. Something analogous to a flight simulator. It is way cheaper that playing with the real equipment. This makes me remember when I learned about networking protocols using Packet Tracer or GNS3.
• Testing environments are really important to provide training for personnel in order to operate a platform properly and to make hacking exercises to find vulnerabilities in  the system. Specially this last part, involves not only the technical expertise on protocols and commands but also deals with the physical aspects of the infrastructure in buildings. All attack surface, (may be virtual, physical) is susceptible to be exploited and used as entry point to compromise an organization.
• Businesses do not wait for communication Service Providers to help to implement security procedures or protection plans. Businesses and companies are taking their first approach to the problem by deploying tests and self-penetration exercises. The network is just a data pipe. This insight makes me think about the role of the infrastructure provider or slice provider to a company... A telecom would care about what traffic the customer has inside the slice? My responsibility as a telecom operator is to provide the resources and guarantee the SLAs with my customer... the same way when we provided E1s, VPLS, VPNs...
• An authorized penetration testing is a procedure that involves a lot of administrative planning! even the presenters (from SODIFRANCE) told a fun anecdote about an "out of jail card" (pun from a Monopoly card. Everything has to be set up properly.
• The approach proposed by the presenter (from Thales Communications and Security) covered a test-bed for a service. I wonder if the same could be done for the infrastructure. I think it is possible, since virtualisation techniques span the different layers of the anatomy of a service.
• There is a saying that states that if the only tool you have is a hammer, all your problems would be shaped like a nail. The key point from the presentation of Franck Sicard is that people tried to apply the same techniques used to secure an IT system to an ICS (Industrial Control System). Every system, service, industry has its special equipment, protocols and processes. The security approach is different in each case.
• The future telecommunication architecture must have the means to provide administrative rights to create snapshots of a slice, in order to provide security features, rollback of configuration and resilience to failures. Could be interesting to think about this scenario.

Faire la pause: European Cyber Week à Rennes, day one

During this week, Rennes is hosting the European Cyber Week. This event, in its second edition, covers several programs that relate to the cyber security treats in scenarios such as connected vehicles, naval environments, e-health and IoT.

This event began with the Journées C&ESAR, which will make emphasis on Data protection facing cyber threats. Conferences today covered the following use cases:

• Naval environment
• Laboratory of research
• Autonomous connected vehicles
• e-health applications and the privacy of the patient data
• IoT
• Government / enterprise reputation management

Each vertical has its own point of view about the treats and the value of its data: different core businesses, different kinds of data generators - consumers, networking requirements regarding QoS - QoE, types of information, metadata and associated value of it. Compromising these businesses would create havoc at different scales: measurements not arriving on time, stealing of sensitive research results; crash of vehicles, liberation of confidential medical data, economic and trust issues among countries all over the world, just to say some examples.

These variety of use cases and exigencies would finally land on the tangible entity all people only notices (the one to blame) when it fails: the network. Oh, well, human factor has to be taken into account too, but it is out of the scope for this moment ;)

The great challenge is to have a complete view, end to end, of all the components that make the service possible; to provide ground rules that provide coexistence and a "pacific" ecosystem. A common architecture that holds them all and provides communication capabilities as the users demand. 

• How will the operation and management chain of command operate all the components and abstractions of the underlying control and infrastructure entities?
• How involved should the communication service provider be in the data management of the segment (better: slice)?
• Up to what extent the communication service provider must comply with certifications such as HIPAA (Health Insurance Portability and Accountability Act) in order to be able to provide services to a hospital? Or an aeronautical enterprise?

I love this conferences because provides food for the thought. A lot of questions, motivated by real world scenarios, that the telecoms (and the rest of players in the industry) must be able to answer. I hope to contribute in this process during the pursuit of my doctoral studies.