 Safety 1
Ultimately, we'll manage to handle large amounts of data, maybe not in the car itself, but outside of it; Silicon Valley proves it. It remains to be seen whether we can handle large volumes that come in very
quickly. The term 'real-time capability' often comes up here, suggesting that there is some kind of tight time frame within which something must happen. What this actually means is simply that there should
be a timeframe within which a process is guaranteed to be completed.
Anyone who predicts that autonomous driving - as understood in this book - along with the absolutely essential Car2Car and Car2X technologies, will become a reality in significantly less than ten years
must be either a die-hard optimist or a lobbyist. When it comes to data collection, both security and speed are important. This calls for standards that, as far as we know, currently exist only within a single
provider or business group.
Perhaps they can even reach an agreement among themselves relatively quickly; after all, their numbers are dwindling, and they have long since demonstrated their ability to cooperate in other areas.
But how many other players are involved when even our roads, for example, have different owners, the federal government, the state, the county, or the city or municipality?
How often is this kind of data actually sent? Do they provide updates on a situation at relatively short intervals, even if the situation hasn't changed? For example, if you're tracking a stolen vehicle using
specific hardware and software, receiving continuous location updates is much more helpful than having to keep asking, at regular intervals, of course, where the car is right now.
Are there any priorities for reports, such as those involving vehicles with flashing lights? If all active traffic is allowed to be transmitted, what about passive traffic? Does that mean every traffic light and
every traffic sign does that, too? If you think about your own experience with computers, those mostly unnecessary or incomprehensible error messages are what take up the most time, just think of all
the emails you receive in your inbox that aren't of interest to you.
A protocol is when you know where something is located in a message. In addition, the message arrives in encrypted form so that no one can read or analyze it, a hacker's dream come true. Of course, you'll
also need a key to decode it. But just imagine the traffic if that has to be sent as well.
So, despite, or perhaps precisely because of, the large volume of data, we need to talk about data encryption (cryptography). The first key concept here is 'authenticity'. You're surely familiar with that, too.
For example, does this deceptively realistic email. which asks you to send confidential information in your reply, actually come from your bank, or is someone just trying to get their hands on that information?
Applied to autonomous driving, this raises the question: Does the system actually communicate with the corresponding software via the antennas installed along the highway as part of Car2X, or has a
hacker attempted to hijack the radio traffic through its antenna? Authenticity is maintained when the partners involved are certain that they are dealing with the people they claim to be.
'Integrity' means that the data received is exactly the same as the data sent by the authentic partner, in other words, it has not been altered in transit. And then there's 'confidentiality'. That is already
guaranteed by law, for example, in elections. No one should be able to see the marks you make on your ballot, so that they cannot pressure you into voting differently.
Of course, it's not that easy to break that down to the radio communications used in self-driving cars. But if you consider the first steps hackers take in an attack, it is to gain access and, through that,
get their hands on some data. In most cases, further steps are only possible after analyzing them. And that is precisely where a breach of confidentiality would occur.
The final protection objective to be mentioned is ensuring 'availability' at all times. Yet hacker attacks against, for example, specific servers or websites would still be harmless when compared
to the 'successful' crippling of entire energy supply systems. All autonomous cars could also be taken out of service. Unfortunately, from a technical standpoint, such attacks are actually even a bit
easier than the ones described earlier, because in principle, you 'only' need to gain access to cause chaos or overload the system.
'Anonymity' is also a value worth promoting as something to be protected. It's easy to imagine that a great many processes involved in autonomous driving would be specific to a particular vehicle.
Regardless of data storage or the requirement to delete data after a certain period of time, it would be better if the vehicle, along with its owner or driver, remained anonymous and could not, for example,
be exploited by hackers in conjunction with mobility data.
| Typical key: file, approximately 2 kB in size |
What does encryption mean, anyway? The simplest method, which works without a complicated cipher key, would be an agreement that goes like this: Shift each letter 4 positions. This
converts the word 'Aus' to 'Eyw', and when there are only 3 characters, to 'Dxv'. You have to admit that not every message is readable anymore.
But there's a catch. If you keep the number the same for a long time and the attacker receives a large number of messages, they can have their computer count the frequency of each letter. Some languages
use certain letters very frequently; in German and English, for example, the letter 'e'. This also applies to words, which provides a way to decipher the text.
However, this example already illustrates a possible fundamental aspect of hacking: letting the computer try out a virtually infinite number of possibilities. In principle, then, no encryption (except reportedly
one single method, which is very complicated to use) is completely secure. However, there are an enormous number of types of encryption. And if, in the example above, you just choose the fifth letter
instead of the third.
By the way, there are an incredible number of optimization attempts discussed even in the 'Neural Networks' chapter. Accordingly, the key factor for a tool that provides a relatively high level of security is the
estimated number of test runs required for a successful hack. No one can say for certain that the first quantum computers won't already be cracking algorithms believed to be secure on a massive scale,
because they might be more than 1,000 times faster than today's computers.
How does encryption work in everyday computer use? Two well-known Linux programs are "GnuTLS' and 'OpenSSL'; the latter is also available for Windows, in both 32-bit and 64-bit versions. It is controlled
from the command line. The only parameters you need to specify are whether you want to encrypt or decrypt, the encryption type and mode, the password, and the input and output files.
Let's start by briefly looking at symmetric cryptography. So if we need to encrypt a message, for example, because the message contains an important password, then all authorized parties must have the
same key, because the message was encrypted by the sender using that key and, in principle, must be decrypted with the same key before the message is readable'
If it becomes necessary to transmit such keys, security issues are almost inevitable. Furthermore, using the same key every time can provide an attacker with an incredible amount of material and time.
A set of keys that are used only once whenever possible is therefore more secure, similar to the TAN procedure used for bank transfers.
Asymmetric cryptography requires two keys. One can only encrypt a message, while the other can only decrypt it. Important to note: The second key can only decrypt messages that were encrypted by the first key.
You send the first key to the sender, who uses it to encrypt their message to me. No matter who has this key, it's useless for decrypting the message.
The message can be decrypted only with the second key, which, unlike the first, must under no circumstances be made public. The only possibility left is that a hacker sends a public key and pretends to be
the rightful recipient. If he then intercepts the message intended for that person, he can decrypt it.
This approach prevents the digital signature. A certification authority can verify that you are the rightful owner of the first, public key. So before someone encrypts a message with this key, he or she could check
with the certification authority to see if the recipient of the message is indeed the owner of the key that was sent.
| 
