Sorry, but this fundamentally misunderstands how sequencing technology works.

The kind of DNA sequencing you do on the population/genome scale is of relatively low quality and depends on a reference genome. It is not a de novo sequencing. Particularly challenging are repetitive sequence areas; high-throughput sequencing utilizes short reads that cannot be unambiguously assembled or mapped in these areas. If you have two sequences that have AGAGAGAGA... on their ends, you cannot determine to what degree they overlap. Only reads that span the repetitive sequence with sufficient margins can be unambiguously assembled or mapped.

DNA fingerprinting, however, relies precisely on robust characterization of these repetitive sequences. These sequences experience many errors during replication that produce high diversity in a population. Your unique suite of sequence lengths is used to identify you. This is assessed with restriction digest fragment length polymorphism analysis (RFLP analysis), an entirely different technology from high-throughput sequencing.

These repetitive sequences are also difficult to synthesize, for the same reasons it fails in real organisms. RFLP analysis is, therefore, about the most robust way you generate a DNA 'fingerprint'.

This situation changes if the sequencing technique used generates longer reads. Thus-far, the economical option always uses short reads, and this technique is perfectly suitable for the kinds of analysis done. It would take a breakthrough in sequencing technology for this to change, which is certainly plausible, but is not the current reality. A future concern, perhaps, but science fiction for now.

This is interesting. But can I know ahead which genes will be tested? Is it always the same set?

Basically, I'm asking: is this really true? Given a drop of your blood I can frame you today for anything plausible for under a $1000? It's kind of difficult to believe.

Yup, it's always the same SNPs.

They can only test a relatively small group of SNPs, and they chose the ones that (at the time) believed differed the most from one person to the next, so as to maximize the probability of being able to distinguish between individuals.

But if you've got everyone's DNA, then you can calculate those SNPs for everyone, and then easily create kits that would allow you to frame anyone you want just by leaving behind samples of "their" DNA at the crime scene of choice.

I feel like I'm asking something you already answered, but just to make sure: this implies, that to frame specifically you I don't need everyone's DNA, I just need a sample of yours, right? And if I understand correctly, DNA evidence is treated with quite a respect in court?

So, if I really can make a sample of "fake DNA" (that will be indistinguishable from the "real" for forensics purposes) for a couple hundreds dollars, I really can frame anyone I know for anything (right now!), because getting a sample of their DNA isn't really that difficult.

It doesn't quite add up in my head, because if it's so simple, why the courts would even consider it hard evidence?

DNA fingerprinting does not utilize SNPs; it uses RFLPs.

if you have the drop of blood, you don't even need to synthesize anything, you just use PCR to amplify the DNA

it's true, there have already been demonstrations of this.

Not that $2 a gene isn't in our near future, but the method you're citing is highly error prone (on the order of 80% misassembled) and (hopefully) wouldn't look like a natural sample to a forensic lab. Commercial prices for gene assemblies are still ~$200 per, so not super accessible yet.