Why don't you build on top of Ethereum? Other projects did so and that worked out well. Building your own blockchain takes a lot of time and effort, which could then be spent on other things like marketing.
Ethereum offers no ledger privacy apart from zk-SNARKS (which is reliant on a trusted setup which needs to run at any configuration change). This basically forces issuers to avail of off-chain solutions which is very expensive and defeats the purpose of 1) decentralization 2) common task automation 3) streamline legal process 4) performances (as trading restriction needs to be arbitrated all the time through third party smart contracts, rather than being embedded at protocol level). So, to be able to meet the requirements of STOs, Ethereum needs third party platforms (like Polymath, Harbor, etc etc) which are basically application-level smart contracts, which resort mostly to off-chain processing, with no exception.
This is also because these platforms have absolutely no say at the protocol level and must adapt to what it is available to them. In that respect, they have a very limited control on the confidentiality level (and in fact they resort almost always to suboptimal off-chain solutions - see for example Harbor, Polymath or even Securitize) and are bound to biblical release time for anything that revolves around NFRs (Non Functional Requirements, such as, for instance, scalability, security, resource management, network configurations, SLAs, etc etc) since Ethereum must scale GLOBALLY all the propositions (meaning smart contracts or set of smart contracts).
Dusk is NOT a general purpose smart contract platform.
It is extremely focused on the use cases it intends to unlock and as a result it can be very specific in the strategies it adopts to scale and satisfy the NFRs of the use cases at hand.
It basically focuses on providing the right confidentiality in the 3 different scenario that render the platform a true network:
1 - Unidirectional atomic transfer of value (basically, payments)
2 - Bidirectional atomic exchange of value (transfer of asset ownership - securities fall in this category)
3 - Time unbound exchange of value (transmission of data stream - such as audiovisual conferencing, content streaming, etc)
We are gonna rollout sequentially the three layers of the network, which will each unlock one of these 3 parts
However we believe that 2 is the most critical as it taps into a market which potential is in the Trillions of Euros
Since we are heavily partial for this particular use case, it is very logical that we adjust the protocol to satisfy all the regulations (and not, like in the cases above, adjust the application layer to workaround the protocol limitations)
To give you an example of what we can do on Dusk which would be unthinkable in Ethereum please consider the following feature.
Dusk decouples the (centralized) KYC/AML vendor process from the (decentralized) security issuance and transaction arbitration with the aid of time-locked accumulators. Cryptographic accumulators are ideal for implementing zero-knowledge white/black listing accreditation of (un-)compliant investors and automated accreditation. Also, expiry mechanism can be added for forcing rechecks. It roughly works in the following way: as soon as the KYC/AML process completes, the transactor identifier (basically his registered wallet address, but it can also be something else, i.e. biometric data, etc) is committed to the accumulator, which remains opaque and discloses nothing of its set. A witness (called compliance witness) is thus generated and returned to the applicant. Such witness is the proof of set inclusion. From that point on, all transactions related to that particular asset class (token) must include the compliance witness. Since the info is directly added as payload of our own (improved) version of bulletproof transactions, the need for open data inspection on transaction clearance is therefore removed. The accumulator is stored on-chain (as it basically is just a number) and recheck forcing can happen through a removal operation which activation can be coded trustlessly and can be as simple as a periodic expiry or by encoding more sophisticated rules through a (zk-)smart contract.
Dusk will host the accumulators directly ON CHAIN, since it is line with our strategy to handle scenario nr. 2 of the above list
Ethereum would never do such a thing considering that it would not make sense to store a data structure on chain that is specific for a tiny fraction of its applications (i.e. Polymath and Harbor) and would be completely useless for, say, propositions such as Omisego, Kyber, Basic Attention Token, Bancor, Status, Civic, Aragon, etc etc.
This means that the applications would have to battle against the Ethereum protocol (i.e. move an eventual accumulator off-chain to the detriment of decentralization).
This would have nefarious consequences for scalability as well, considering that these proposition must resort to continuously calling smart contracts in order to arbitrate the transaction operations (to the detriment of both performance and privacy), while Dusk is in the liberty of enriching the transaction directly with zero knowledge data from the transactors (the compliance witness), leaving absolute liberty over which vendor the issuer wants to rely upon and keeping the centralized compliance-related operations (like for instance KYC/AML) separated from the transaction stream.